V .. V I .V. .1. n a. v V... ... _ ..0 ~0- .' . .00 0 . V . o .
. I o . . 401 . _ I. u v I | . . 00.0 0.0 0 V 0 . . m. 70..
. V .— . .V / ‘1‘”. o WMAJ m‘fihgfi LO... .1i‘4flu {Lu-o0_ 0. 0 \pruv1.4.0V 1. (”0.- av n .00....00‘0. . 0:. 0900. “J..-3.00.1123.- .00 . 7....- 0 “0 n 0) 50.0.1 V V.0..00..0.040.0...vu4.0..0.| on... .V 01“.?” .0... a .3010 . 0r 0.. VV. ... 00....V.VVJ—1......0\. VV..V.~4000H40’0.0.00H...
0d . .... 0.2.. V ... .... V. “and.“ 0.10.4... ... . V0....0 .... ... . Vv. ... co. .... .V ”V. ... .o .. L. ...... 0.... . 0.1... 1910.....-» ...-a. u. .. .. .....IV8VV : V.. .50. . o. 0 . V 9. - c . ... . I. . . V . 3V... 1m . ...VJH. V
0... 00...“... . .00” J.V.0.. 0.0.. .40.}... 0' .I 30.09.10.074... .0 V. e. ..uén. ... V .00 I o . . ...) 4 .....0H0 2 . V -1. r.) .03... !.)v.0 .- ... ....Q'Munl.
r0“..0.0 V . .00004000 030,104.10. ...0. v .....I 10.0 o 0 .o .0‘0 ’ 01.41. ,4.- .. 0 0 .0000 . 0. .0. . . on...00....:.00\ o. ..00vo. .00n'0! . O. .H.V 0 0 1 0.. .. .. .... V0.0V 10-000.“... . . ...0 V 0. .0. v 0V V .. . .00“ fimmu
1.3 4 than“... .C ....4 o .ngu . 34.0w. 4. 0'. 0.1.... ..v 0.0... . . 0. ... 2003.0 V... 0.. 10.0 .4... 0.0 ... . 4.. V. . 4.. . V ’L .... . .. . ..0 300.0. )1 V 04.0 V! 6 . . Mia-no.1.“
90.0%”...411111... 07.. :00... 1...: o ..0 .71.... \.-\1.f 0 ... 2.11.. I. .....0 . .V ..V . 9... . .. .V o .....0. 3.00 . . 0.. V .0». I V. . 0.0 .4 . . ..... .. . . u n V VI . . ... ... 4.0.
on but“! !. 0%. 01.... 11.... .v . 000.0. .0. V001“?! “0.10 . 0.00 s .. ... 40. 0 .. o 0 . - V. ... 1... .00 ...-“0
3L80.u..0.00ul.$41.3. V141. 1.0K»...t . 03-004.. :0... ..00. .. . 09.00.. 40.00 ... Va... - .040 ...-000 .- 31:1 Tr
“$4.0”: 0.. .00 . .0. . .wv .. r. ..J..010n~V ..r‘ip‘ .8 .... 0.. . 30... 0.7.7. ‘0'! 03.0.0.0 671.....an ...1v490. ...
T485 351;... ...V .51.. ....V +1.... Jeanna“.........huVVnpercVJ...”9........VV.. . ..V...
0x 0 0‘0 0... .. . .4 . p . \ 31:9 .0 1.4.1 V. u . . V . 8 . .V V .
.H' .70.,» "my $0.31.. 1.50... 00. Q4 4 ......Om’o‘ . .0 00070. . 10. c. .. ......0000m.5....0..00 . .. I t o .0 ”020.00.. ..0 00 .0: 0.\0. Vq. . .er 2.0.. 9.00. ...0. .. u. .. .. u c 00 003.00
0. ....fl' 0.0....
.0.
.- 4.0. .00
J..-p. ”0%. .4104... .0 ..v. ‘00 3:11.? 0%"5. ...00.04.0.... V... .40. ...-000.7... 0.. . .0.“ 0.9...
0-0..- i. Q: 039
’1’ 00' :1'40NI0QH40V'uuVi4-0’
.0: 1.... .
...0 ..........1VVV..V.51~:!
..01IV.0V0.V.V.0.
.04. I. 0 VV. . 004.. .....V. ...v... ““91““... ...n... u .v .V . . I a 3... u." ...-939. V V V . V V .. .1111»
'f. . X0. 309.0000" 0..0.1.|0. 0.0.4 . . V . V . V . . .. .flP..¢.....f00...Lk.-*o.61
...: V... 0. :0. ”.... ... . ...... . 03.8.0: ... 0 ......~ 1.... . . V V. . V I... 4. 0.. .
...: ...! ...v .
0.... .V 00.00? ..0 0. 0.
V .. . 0..
....f‘fl 1...!00‘...0‘:..-0.0 .100
.0 0 a I. o
1.01 ... 0.. 9000- 0 0H0. ....»010 . 4'00. 0.0? |.. .604000“.V.. 0....000
0.0300411 101000...
V V . .. . 0. 00. .4 .. .V . V .
. V... .. :1... .30 .....no... ..1). .0. ‘4- ..V. .m...’ 0.47.104.- . .0 10.....4. V .‘ .1 I .... .4. .. V. V run-H.040? 1v? 0w“.f‘-V 0”“.8
......31.1... t ..V t ...... ......o. 4. V. 9.0.1. ......0... .. .... ..v 3 :r. 3 ..0 IV. . V V ....9 p! ......1 ... . East...” 14, ... 0.... J..
. ”331.10F01’4. 001.0. IgUL 00.4...11 0 .... . .0V . .V . #0.. V. V K ....0. . .V w ..l ..0.0 ...... .01.. 0.. $10.14 .00 .v.
V .00 0 11-00'00‘0. I .V000§0V~.v!.h.u..,u.'010. 1. .~0.0040V.-N..0 V. ... 1.13“.“ “v 0.. 04 . V «in ... uwfluhqhomb‘ag
(0 I V V. .. 0 . . . I... s V
... ...4 .. hauntnflaiv ...NV; .... .50.}: ... . J3...» ......i. 0.... u... . .33.:
“won-1V1. . 000‘.“ .0.- VNTQIHM. ...3. . .. . ...
p .
....00...‘0..0‘00 00.70”“: 0VW3V0+64H0 0.040.. .0. 01.0 V. 0 0.. —
I 0... s
a. . . g. .0."|‘.I..00w 0 00.. 00.. ....
V . 31.... x... .o.’ .... 1.1.4.
“ . 3...... :23. .. s .
”00,0..Mn‘ ‘00.. .440. .0. “W000 .0. 0
r0. 3‘ .100. 0‘0.
0V .0 c 000‘ 1......
. ...:MN‘0 V “xii; OOH-bur,
V V... JV». flan... .. ......VVVVVy ”mu
. ii?
hm. .Nadmo“
V V 0 . ‘F 0.»- .‘000. . v 0400...|: .. 0- V... V . .. . .
a.” .du...V0u.n.V..uH..H-.olq011:" 0.114)... 4b.“... a 4.1.0.. \.4”.&uu ...... .0" .. “.4. . . ... . l0 .040 .0100 ...... ... V. V 0.... .o ...-.00.... ... 0.. .8....m.. o V... . .V . V V V. V V . U318?“
".i’ 0000... ‘n03oer: 4 10". 0 . .0 4 00 04.00.00.0‘4’01’ . . . .VV1X”'0. 0.0.. V . .0. . . V . .... 0 o .0! .0. . . V . V . a . 4 4 0L
4.. 0 . . ...-W... {3. . .0. 50“... 4.03.5.0. «IV... ......V .1. ... .1.r.. 3.0... 0 ...... . . l. 0. V .V .. ... V . . V V V . . V V. V V V ... .0 . .0. 11.
«V. .. .0010. .0104. I: V": 13.03.00.594th .....0..’..u0.u. in“. .0 ...... . V .0. ....1. V .
. . . . 4 i . l 00. 0 0. 0.
4a.. 0.0“: 000413509! ’0 .1 4... I. .0.
0. V. .
13:0,.» V. ..00. .0. 010....
1...... ...WV. .... .311:
0 VV.:...00V0.V......
. .V 0“ ...00. ...]: N .00.. .IVVoJ
. 0 . 0.
I. ...-0.4.00.0“.0uVoVQV’Ibfl. .4‘..K.QV0.H.
3 .o 4. .4 ..I’IH
a
. .00 .00 .0 . .
... 0.. o. .0...0.. .- . ... ‘00 .0004 0
.l.‘..00. ......o. . o . 0.. .
r. 00. 0.102.000 V.
..1...9.34..Iv..fi..u 0... V..
o 0. .
.‘ .0 .
.ufl.k0.iu.0.4v. 000 ......0 040-
V- .-04‘0 ... 1.5“. 0.00. 0.. ’0.
..c. 0 .1000 0a 1 0 ...110.J.0..0.04 .. .
90”.. . 0. o. 1 0.4. 0 0 V. '00.... 0001.10 4.00.8..\ 0043.
0....0... .9
ctr-......QV
. V0.....V.0.0.
..r ...0 5.1071110 00.2.0.0 1001......0108... $0000 -..V 08.. V... ... . 0 ...0 0.... E V 0V . .
0000.. 4.04 I. .V 00. 01.11.010.149 .90. .
'1 .- .11 0.4 .... 0.0.71.0... . .404 a. V .0.V00V4.0... 0.00 . V. .00 " 0 B.
It. 0n.."0-0..0V.’|V-0.4.0 0. 300,0 0.. . 0.1 0.90.... .00 c. 0 .004. . V V . ... . . n V. v V . . V 0%.). i104. ‘0 590‘... I
V V . . ...-V Vv . .... . .V V . V V V V 13......0.“ 3-02....) ....
. 0: ...... 0.0.0.04. .. 4 V... . 4.0V 30.. . .-. V V. . . . . . V . . V V V 1-434...‘ 00. .0‘0‘004. “a. ‘3
V V V VV . . V V V V. V . ‘8‘ ’13....104. .1 I
2.....0 n . .. .V . .1. 0‘. 30!...
.unuerV .0... 1-0.- ......89 .... vhrmV .0. v 1.. 2.0.3.4... .04. 10...“... 0. . 21.0 o 4.. . ..Uoou. v. 1. 0:0 .4.) 31%., 101.4031. 04g..‘..V 0
. V. ..0 n . .T... .V ... 01.3.0.0. 4... 0.004.... .3I1Vut11400‘001“
..A .0. .... .0V. 00...... ......000......00 _- .00. 0.7.0.5.... 70:01.00 0.. V . 1.4;". 0
00).} “0| 04 00. 00.. 0.00 ...-0 00. .. .V .00- K . 0. x ..‘V-u. 0.... .0\ 0.0") 11'.
......00... .9 3.134.013... .... IV! 3. . V . .0820... ... .fhw..00.w.0..9~.. .....Vum .... .. ..hun L . . . .63 0 . J.1.WU.\9.m\.-...0 ..an 0.30.0.“
. V- . V. .... tr... ”hugs”... .222. ....V
0.1 . . . V V . . V V . - . 00
. .4. .0 0 1.0. foam-V |.... . V V V . . . V . V .V . V V V . . V V V . 6.0.0.3100. "0.: 0,333...) in
0 r..u0up..-..Vu..-V..VV .. 0.. 00w"...V.V. ... ....V V V . . V V V V . V V . .. . V . V . V V . . V V V V V . 1 VV . V V V . . 01.04.0033,]; V34...0.1.0ro0'.du.l
0w. . I... ~ 1... . V . V V, .V V V . V . V . V V V . V . 7.0m-‘al..00v‘..0.01._000(?0084 011-0
08...... 3‘.’0..”.M.*V
V. . .343... ”nah.
V...' 20541:),qu .0.
. ......
. V. 4. .400; ‘9. .404 10.00. ...)...0 ’30-‘01. V
\ 0111.. .. ... . . ... .. '34:: 0o ...0. 5.1.0.131, . 9...
.l.....~..4..0.41-n. .00.. 11.0....- .w. .0? V0.
. .roo0d .- 0... .0 30.0.0 0.2V. 400.1...0 ... 0 00..-
..03. ..‘.001.0-0.. .0.
V...-.. .040004
1.1003} 0|. 4.... 0: .013”... a V .V .V . V V . . 0.00 00.40 0. 4v '0 00 p.
. . . . .0 . .. ... .V n V . . . . V V V . V V . . V . V. . r .0... a». 0-. 31.. u ..000. 0: 001V
0.. . o .0 ..0 03...... ‘1... .. .... . ... V . 1.. r . . 1.4 . . V . .V . V . V .00.“? ...... 3. 180.00 I!
3.6%.... 1|..t1..90§.... on. . .3 .0 . ...... .V .. .0 . .4. 0V V. . 1.1.000... .oV0...n.014V44.4VV. ..0. ... 1 .00
. .....h. ... I. .4. . V V VV . 0. 1 - 0.. .0031? .E
4 o .m . .40.. . . . V V. V V V V. .V V V. V V V V . V . V. V . V 19......0. V ..1 \. “MA-000:”. 0uuwudnflvunn4"..00 "an: .0..o¢I-'Hu10
V . V . . V . . . VV . . V V V . . 2.56.. 000 . . .V.V0... 0. so]. 1..)0.0.1V.V”0V03 90V “gloat.”
V 0 004 .400 ..I. 0. v ... . 0.004
. 0004-. . 4. '0 000.”! anon-.040 .... \0_.£.o0.l4‘o 0 1.....0.V.r0”..' 40.0
.011... .4. 0.0 1.1.0.... .. ... 1). .V .....o. ...
V 9 V
fin»... V 0.0.... 70 V . V . . V V V V . 10.00.0001 .V
V 00-00.."00M 95.400007...VIC1 . 10.0.; 000. 00. V . . . V . V . V V V . . V .. V V . . . . . . . I... 00.791340..- 0fl.UI..-l' {'14:
5.20:... . . . . V . . 1.04.0 . .V ..00. ... . .V 00.1. . ..< 00...... 0 2 .0. . ...Yo. . 00...? .....V . -....1. .V V. . .V V . V V V . . V . . V . V .. . V . V . .V V V V V V V . . V .00.“; 4.40.0..3..|3..0.“~:0'0‘4X.0.0d
_ ... P... 1.00... ... V. 0 . .. u. ... 40 . 00 .000 .. V. . . . . V V . V V V V V V .0...‘..nolnmuq.vfi.\1. .04.. n. .00. $400.01..
«.5. .0. ...: V... I ... .000: . V . ..l..r'. .... 0‘01””."104 . 5 040 1.10
V I 0. 0.0. .. 3|Vr..._ooV1. 01,0.x 0 40 .00 .‘X . . . .V . V V . .V V . . V (‘0 ’0. ‘10!)0.0-’040 “0-00..
. 0|!) .4- 4 . 0]. . .0. I... .0...q 0 0| u’Fo‘ . V V V V V V .. V V. . . . . V V V V V . V V . . V. .V . . V V . V. V . . .4 01.03910. 7.0.30“... “”003 . 11.0.Q1. 00. 0.0
000.10.. I. .V a. " .... a . ..0. “0““..03 .4..... V . V . V 00. . . V V V . . V . V. . . . ... 4.0. .05. ....00. 0.0.2.0.. 0.01...0‘u000.-.4V.0u0.3.)bM-.m0.. 110*
0.. 5.01.0: . on. . ...0 V l .0. 0 T; . ...0 . . . V. . ...V . 9 . 090.0. . ... ... ......0 9.....13.’ . 901.40.05.34...‘ 0. .. .3050.
...n......~......:: .V ..h...n....r...u.n... .. V. 6...“: .V V. a .. V V V. ... _ . . 3. .. .....V. .. VI.V.V.|VV.V.i. ... .... VV. tun.-. ...).V V. .... .
. . V..\ 0- 0.000 .0..- 031.00.. 10 .005... . ... . 0 V .
..an V . V. . . 0 V. 4.40....‘nl..’.0..0.VV0V4....VV .031}. «0..
V . ...... . .4. «. .0..V . 1.. V I 0. {9.90.0-0 0.... V. ...n .0 04.0.1004 50.40 .1 . V.. . 00. V 0 ... . 00. . . VV. . 10.0.9.1!) .00 0.11004 .400 ...0-
V 21... V». 09.93. . ...”... ....Z. V o ... V. .. V.1DS.‘!......0..V
0.....2 .10.... .... V .00. .. V. . V .
0..0...o... 0.4.. . . . 3.0. ...... .V. ...-v ... . .1. 4.4. 0 V. . ..0 V .000“ ......n!..0. 0.. .
. “01.1003... 0. . V V... ... ..I. .V V .4. .v. I... -0...V.00Y $0.0 o... 0...; 0...: . V
.‘0H.0.‘.. d ... 0. . I 0.. 0 . . (09.. 0. .0 . .. 1 . 0 . . . no. 0 0' . V| | V 4.. . V V . V .. 000.00... .000. . V
V. 4 .V . V40V. ... ......0 .4 V. .....1‘ .0 ..1. -. ... 0. 0 V0. . . V . . . V V .V.VV-0VV0..00...
. .. “1000.14 .0. ..3 .0.. 0'0 2-16.". 0.3 .40 0 .400 01. 0.80.101! .. . 0. 0......- .0.....V.V .".... . .V. . . o. .l o 0.... . ...V . . \ u . V .. . V l ....
.0.00V-. . 0.....0. . . .. .. V . . . V . . .90 41.0... 9‘ ...0.V.0.0....V
..00!.. 0. V .. V .V. . ..00 11410...
.0....?.30.0..V.. .. V V . V V. V ...!0u5011
v 14‘...0v......o1..0
.VLV.I..... 0..VV.
V. V 0
.....0? .30...
.70 .0
V .00.. 0 0.... .4
v 0 '0‘0 ... O.
'0... 0. 4.40.”...00’:
V . 0 .V 00.
. .0 0V. 0
. . . ... .000 ... . 00.. 0V 0V. .0 00
. V .....0 .. .00 . ......'0 ..0409 . . . ...
. . V0... . . ... V
V .....0 ...
. ...... ... ....VMPEL
V.. a ...00 V V V
. . . V .V V V V V V V V. . V V V . .0... . .
v.0 . 0. 0. 1.0. . V. V. V . . V V V ...—0 {00.00.03 0. ~.‘..l..0v-’1.”.‘0VV
. n V V . . V . V . V . v4V-.V ... 01V V .
.0 V.. ..v V. . . . V .. .I
.. .0. 0 .0 V ... .. .. 4.0.
. 0.0 0
431.000.. 0...7.000.......V0-.V01..0....1V_.......0..v. 0.
0.
’0 0. .-.
_. ...
. C. O I
0 L. ... .. III‘..‘.0.. V 0. .00..-“ .0. u .
.4 7.9.00. V.......... 0.... V .VI... 4‘. ....0.. 09 o.
... o. .0 .. 1.1.0.11. .. .... 1.0.4. V. . V V 0. .V .V .....1. 3 V ...-V 0 .
0 . . . . . ..- . 0
.....0.V....0. "004.4..0014 . .. V
V.0..... l.- 0.0. 01.. I 0
V V. . .
0 V . V V . V. .0.V.0...lo.
I. 0 o . . . V V V V .
.... 9 V 0.... .. IIV‘. .- V . . V V V V Tim’s-I.
. 13.1.1.0 . V. 8. .
... ~1..1’r00..§.. on!
'04! 00.00....4VP.0 0...
. ... ...... 2.
. . 04
.0 . ... . V .
0 0.- .— u . . 4 .
.04. .0. . 0. .. VV V V . . .V .. V....0V0V0..|V..0..V ....V
00 .-.. ..v .V0. . ...... V . V V V . .V V 0.. 0
V 0. . . . .0...-
V.V4\000VV .‘ i V . V
. o
0. . 0 0
0 -
F u
.00...4 ..0 .. V .
0 4 . V . . V
I ......V. . . . . . V . V. #810.-
. V . ..V. V V. V V V V. .L4.0Vr0‘..‘00.
04. V V... V . . V . V .. 0. .... Q
.0. .0 . . V .
. .V V...
... .
.IV.‘ .. ...... ....s .V
. . . VV V .. . 0. 0 .
00V .0. ......101 . . V . .. V . V \V . . . .
. V V V .0' .1. ’00
.
01-0 .0
.0. .. . g
.0. . .00
.- o .00.
0 ......-
u - 0..
0 0
'. \0
.00 . 0.
4 O
O ‘-
V. M... : .. v .4..0....|.V V . V .. V s .. .
V V... t... . ...o. .00.. ...... . . . V. .
V..'. .VLVI.V0V ..0 30.10.3400... . ... . 000 .
V . .V . . . .. .-
....0.. ..0.0 f. ..Hhu..... .4 .I. 0 I. 0. .-..0- . .n .
. .J...01VV..0V .. . . . 4 V .
. . ... . . .0 0.. 0..- ... 00
.... u. . . o . ...00 o .0000 L
O .0 . ... 0‘s. 0. 0
rm 00.1.4 00 ... 0 . .
. I. . V . V . V
.0... (M. “u..u0.0....0)>.4 0. 20.00 ..0 at ..’..1.V.v| .04 .V ...M ... V. V. ..V. . .000 V .... ... V. .V I. .v . . V0.4 . .0... ...
.. . ... ..0 ~V.I. .00.. V
. ...-V . .0. -. v01 0...
4 0 ~. .9..0000..4v..
. V..0.0V \v.0..0._. .. .. IV .V. .3... ... 0V. V V V ... I V 0 . .044 . I...
. .
.0l .0; 9... . V... 0 0f... . ... V.... .0 V ...... ..l c 1.0 . V . 0 .... . o. ... ... .5 V0 .1 .. . .0 . . .o
.0 .V- 00. ‘ID. 0.... . '. .. .... . 0.2.. c 0 ... ... V 0! _ 0 o v V
V . V I 0. . .
. V V . (Vi... . ...;04. .. .0. a I I 2 .V.‘ ... V.
..0 a. J..? ..I0uu.l..’.. .0 .. .. n .0 V. . .... .. ..0 .. . . 1 . . ..00 . .V r. . ... . |. .0 no .. I. . .
. . ..011 .. . V... ....V.. . ... V .... ... v . 0300.0 . . v .... . V. V. . 0 . ... ... .... . . .. .
0.158.010.3LV. V18. .- . . . .0. .- 0. n 4 X. 0. .1. v. 0... -. \. . V I . V. n.
...V0.V.—1043 V. 0.. .. 0 IV. .004 0.0 o .07.. 1.2.012 . . . ‘ I .10.! .. - .0 0 . I c . a. . . (0V .. ..
012.91... V .-.. V V .... . I .. V V .....V . ‘ .... . V - ... . . :0 V .V . ... 0. o .
.I v... 0.40.. 04. .. o . ....u ..0 .I ....o 0 ...... .0 V ... . V . . I . . . v .... . u 4 .. 0 .. n ... . .0 u. .
. .00.... I. V V ... V. I .4 . .. V. ...: .0 .. ... .. V . . V . V. . . V . .0v 0
. .0 0. . o . I I 1 0.. .00.; 0. . 44.. . .V .0 0 l . . .u .10.. 0 . . 0 V .0. ..V . ...!
.. 0.. .... .0..V..V . . .9 I) .. V. I 04 I . 0 . ..V . . c
. .
4 . .04 u u 0 .. 0 0
v 0 _ 0 V I 0 ... .0 0 I . . V V . V ... 0.0.. 9 V 4 . . . 00V
. .0... ......v. . oV — 0 0 . V ... .. I . . V . . V .. V i L. o .. 0.0.. I .0. 0 .... . 4.1... 0.00: I .... n v . u . . I .
n... I :00. V. . I “.... V . o ..0 . V :0 *\_ o g. V .— 00 V. 0 . . .... . 0 . .0 u 0 l ...0 . V v .V o s V .. .0. V .:.¢ 0 04.. 0 . . o . ... ..0 O. 0. 0.0. .4 . . . .. .n . 0.. .u . n .0 . 0. .V 0 . 0 . .0. V ... 9.0-0 . V ..- 0 .00 . V 0.03 k. VON-Mr...0.u.. V00... ..uu0u ..w 0 . . 0 0‘ 14 . 0 . o u o .
...0. 0.0:... $7”? 0.. V‘.. ... .140 002.00.1000... .I .914 .. .... .4 .0 O . v 0. lo... I . V... 7..- . .. V . . _ . . . . o. . .0. . .4 ... . 0 ..0. ... 0. . ..0 ._ .0 ..01.V~.V0..0..004 ...I . .0 <00 0
— .0; 00.. V 0804 V ..v...0.4. V .n". .100 .V: 0 c ... . . 40 . . . . . . . . V00. . 0 . V
. 1 l 40 V 0 ..0 — 0
I I - 0|. 0 I
I'. I
r
W
!
'
2 LIBRARY
110001 Michigan State
University
This is to certify that the
dissertation entitled
A MICROSCOPIC HYPER-SPHERICAL MODEL
OF TWO-NEUTRON HALO NUCLEI
presented by
Ivan Brida
has been accepted towards fulfillment
of the requirements for the
Ph.D. degree in Physics and Astronomy
(rrwajar Professor" s~§g\nature
EJ434011 Z 00%
Date
MSU is an Affirmative Action/Equal Opportunity Employer
PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.
DATE DUE DATE DUE DATE DUE
5/08 K:/Prolecc&Pres/ClRCIDateDue indd
A MICROSCOPIC HYPER-SPHERICAL MODEL
OF TWO-NEUTRON HALO NUCLEI
By
Ivan Brida
A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Physics and Astronomy
2009
A Mimi
\I-I' llin'i' ll
morpwrutm I
E‘ijilaifih iii it
Gaillfifilitln. I‘
am mm in nu
hi'pt‘ir-rmiiiil ;:
are? usml fur tl
In the“ pm
6H6. Tlli‘ (rm
Eil'liliiii'ili is ”N
mule-ls and in
1Hi}. railll itliil
dwriptiun HI.
Ffliiiluwl in t
Crucial ml
h
illxll
arr: liwn ext:
I'I‘iHii‘Iliri'in II
ABSTRACT
A MICROSCOPIC HYPER-SPHERICAL MODEL OF TWO-NEUTRON HALO '
NUCLEI
By
Ivan Brida
We have developed a microscopic cluster model of light two neutron halo nuclei that
incorporates the few-body asymptotics in full extent. The wavefunction of the system
consists of a core and two valence neutrons. The core is given in terms of correlated
Gaussians. The three-body dynamics between the core and valence neutrons is taken into
account by means of the hyper-spherical functions containing an exponentially decaying
hyper-radial part. To avoid the spurious motion of the center of mass, Jacobi coordinates
are used for the entire system.
In the present work, the model is applied to the lightest two-neutron halo nucleus,
6He. The central Minnesota nucleon-nucleon interaction with and without a spin-orbit
addition is used to bind the nucleus. The results are compared to those obtained in other
models and to experimental data. Basic structural observables, such as binding relative to
4He, radii and one-body densities are in agreement with other models. The microscopic
description of the core allows us to test the efficiency of Pauli projection techniques
employed in the few-body models. We demonstrate that proper antisymmetrization is
crucial to bind 6He against three-body break-up. Overlap functions between 6He and 4He
have been extracted with the aim of reaction calculations involving 6He. In particular,
two-neutron transfer reaction p(6He, 4He)t at 25MeV/ A is studied.
I intuit
with mt“
()II'II llil'ti“ '
$97311th
‘Ilil “ll I:
nae-Int-t-rs l
Grrg'T‘ ll.
I \tuiiilil
annual .\ln
I’afii‘ifinimi
Xi'xi. l
Stair lilill'i‘
mt‘lll ii! iii:
Whirl fail a
the Dignim
lam grei
iii-r» Hit i, _:
Pitts rm im'
riililli’x'w] it:
and Angel”?
dim. llll'liti
MIMI). and
LN lJlli
ml ba'lJTlii‘rg
ACKNOWLEDGMENTS
I would first like to acknowledge my advisor Filomena Nunes. I admire her patience
with me—a stubborn student of hers, the freedom she gave me to develop and work on my
own ideas even when she did not find them useful, and her guidance and jokes throughout
the years. Being her first graduate student, I hope her next students will be less stubborn
and will give her less headaches. I would also like to thank my guidance committee
members Edward Brown, Sekhar Chivukula, Remco Zegers, Vladimir Zelevinsky, and
Gregers Hansen for their comments and suggestions.
I would like to thank Robert Wiringa and Steve Pieper for discussions on the vari-
ational Monte Carlo method. I thank Kalman Varga for discussions on the stochastic
variational model and for providing me his computer codes.
Next, I would like to thank the Department of Physics and Astronomy at Michigan
State university, the NSCL and especially the theory group for providing a great environ-
ment to work in. Special thanks goes to Shari Conroy without whom the theory group
would fall apart in chaos. The financial support of the National Science Foundation and
the Department of Energy is also acknowledged.
I am greatly indebted to the other members of my cohort: Matt Amthor, Jon Cook,
Wess Hitt, Andy Rogers, and Terrance Strother. I will never forget all their inappropriate
jokes on my behalf and the long nights we spent working on homeworks. Special thanks is
addressed to my current and former office mates Jeremy Armstrong, Biruk Gebremariam,
and Angelo Signoracci. There are a lot of thank yous for other graduate students and post-
docs, including Ania Kwiatkowski, Giuseppe Lorusso, Rhiannon Meharchand, Michal
Mocko, and Arnau Rios.
Last, but certainly not least, I express my gratitude to my family, my parents and
my brothers for everything they have done for me and for helping me to make it to this
point in my life.
iii
:il
“7“
Contel
List of Till)
List of Figi‘
l Introduciiu
l.l lldliiili
1.2 Tli'iI-ni
1.3 OU'H'li'
ii (iv-m. '
l3 Mullml
1.6 Outline-
2 Valence [)éil'l
2.1 (‘llllflllllw
3.2 Ollll'fll.
2-3 ”Lilli I.
“2.3.1 l:
2.3.2 1'
2.3.3 [5.
2.3.5 (2
2'4 b“Mum.
2.1.1 1,...
2-4-2 Ill"
2.4.3 it:
Contents
List of Tables ...................................
List of Figures ..................................
1 Introduction
1.1 Halo in nuclei .................................
1.2 Two—neutron halo nuclei: 6He and 11Li ...................
1.3 Overview of 6He and 11Li: experiments ...................
1.4 Overview of 6He and 11Li: structure theory .................
1.5 Motivation for present work .........................
1.6 Outline .....................................
2 Valence part
2.1 Coordinates and bases ............................
2.2 Other ingredients of the three-body model .................
2.3 11Li in the three—body model .........................
2.3.1 Introduction ..............................
2.3.2 10Li ..................................
2.3.3 Interactions ..............................
2.3.4 Results .................................
2.3.5 Conclusions ..............................
2.4 6He in the three-body model .........................
2.4.1 Introduction ..............................
2.4.2 Interactions ..............................
2.4.3 Results .................................
3 Core
3.1 Stochastic variational model .........................
3.2 4He in the stochastic variational model ...................
3.2.1 Interactions ..............................
3.2.2 Results .................................
4 MiCH: final assembly
4.1 Core and valence together ..........................
4.2 Variational Monte Carlo ...........................
4.2.1 Monte Carlo essentials ........................
4.2.2 Can we trust ourselves? .......................
4.2.3 Wavefunction optimization ......................
iv
55
58
60
60
66
74
81
2.31
- 0‘ I'v' Jun-In: ...-—
0‘
Summary
6.1 531mm.
62 ”Mint
Implement
..‘11 Ltii'iili
.12 (Ila-Int?
Further t0:
8.1 .l‘l’lli ill
8.) {\‘IIIIIii
C Comparat i‘
D Wavefum‘ti
5 6He in MiCH
5.1 Antisymmetrization effects in 6He ......................
5.2 Converged 6He ................................
5.3 Overlap functions ...............................
5.4 Two-neutron transfer reactions .......................
6 Summary and outlook
6.1 Summary ...................................
6.2 Outlook ....................................
A Implementation details
A.1 Local representation of wavefunction ....................
A.2 Operators ...................................
B Further tests
B.1 Triton tests and the story of bad points ...................
B.2 Additional tests and checks .........................
C Comparative optimization on two independent random walks
D Wavefunction normalization
Bibliography
87
87
93
101
113
118
118
121
123
123
125
127
127
132
134
138
140
r 31
(\Q\
Pi"
(1H!
Stir]
Kilt
Slim
PD i
List of Tables
2.1
2.2
5.1
5.2
5.3
5.4
5.5
Probabilities of dominant components in the T Jacobi basis in 6He . . . 43
Probabilities of dominant components in the Y Jacobi basis in 6He . . . 44
Combinations of quantum numbers for K = 0 and K = 2 channels . . . . 89
Squares of the Raynal-Revai coefficients for K = 2 valence channels . . . 89
Energies and radii of 6He obtained in different models and from experiment 97
SpectrOSCOpic factors of the five dominant overlap channels in 6He . . . . 105
Probabilities of overlap channels in the 4He + n + n decomposition of 6He 110
vi
*2 5'
RM
List of .
11 Lower gm
1. N1 .
L2 bfiiiillt1.i
v-
:1
n.
‘I .. '
.11 .I‘tttftiliiiili
c ‘
it'll'illllit‘
..3 Hit-"11118 iii-
.4 T1161HW'.‘
:3 Di-ptiis [at
2,6 Bimiinu v:
2..Binr1ii.gvr
Dt‘pi‘nill'llt
2.9 T111T‘t‘-l)l H1
2111 R3125 riiii'li
2.11 Proliilliiiiti.
2.12 Dr-penrl
t‘lu'
21'?
2’ HYDPHM;
,7
'IdI‘UbI ('(Irfl‘
‘ dlil'
d1 (li'Iii
List of Figures
1.1
1.2
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
2.10
2.11
2.12
2.13
3.1
3.2
3.3
4.1
4.2
4.3
4.4
Lower part of the chart of nuclei ....................... 3
Schematic representation of structure models of 6He ............ 8
Jacobi and hyper-spherical coordinates for a three-body system ..... 17
Coordinates used to define two-body potentials in the three-body model . 23
Radius of the 9Li-n interaction as a function of deformation ....... 32
The lowest energy levels in 10Li produced by the core-n interaction . . . 34
Depths of the fitted 9Li-n interaction as a function of deformation. . . . . 34
Binding energy of bound states in 10Li as a function of deformation . . . 35
Binding energy of continuum states in 10Li as a function of deformation . 35
Dependence of the three-body binding energy of 11Li on Km“ ...... 36
Three-body binding energy of 11Li as a function of deformation ..... 37
Rms matter radius of 11Li as a function of deformation .......... 37
Probabilities of structural components in 11Li as a function of deformation 38
Dependence of the three-body binding energy of 6He on Kmam ...... 42
Hyper-radial dependence of dominant channels in 6He ........... 44
Jacobi coordinates for a system of four identical particles ......... 50
Radial dependence of the central Minnesota force with u = 1.015 ..... 58
Convergence of the binding energy of the MN and MN-SO 4He ...... 59
Effects of correlations in the Metropolis algorithm on local energies . . . 76
Flow chart for the decorrelated Metropolis algorithm ........... 77
Effects of decorrelated sampling on auto—correlation coefficient ...... 77
Effects of bunching on data from Figure 4.3 ................ 79
vii
al
4.5 II
a"
C
4.6 131
5.2 ('t
5.3 Bi]
5.1 1‘2,
5.5 D i:
5.6 113'
5.7 in]
5.9 .-\~\
39 ('tif‘
I
5.111 I11“:
1,...“ WJ '31 1 (‘fi i:
8-1 Pitt}
82 Bilii
Ilili‘lfll’fi in
4.5 Running mean energy and energy error estimate .............. 79
4.6 Block values for data from Figure 4.3 .................... 80
5.1 Three-body binding energy of valence channels as a function of po . . . . 90
5.2 Convergence of the three-body binding energy of 6He ........... 94
5.3 Binding energy and rms proton radius of 6He as a function of po ..... 95
5.4 Point proton and neutron density distributions in 6He ........... 100
5.5 Point proton and neutron density distributions in 6He in linear scale . . . 102
5.6 Hyper-radial dependence of overlap channels in 6He in linear scale . . . . 106
5.7 Hyper—radial dependence of overlap channels in 6He in logarithmic scale . 107
5.8 Asymptotical hyper-radial behavior of K = O s-waves ........... 109
5.9 Correlation density plot for the ground state of the MN—SO 6He ..... 112
5.10 Hyper-angular probability in the K = 2 s-waves overlap channel ..... 113
5.11 Cross-section of the p(6He,4He)t reaction at 25 MeV/ A .......... 116
B1 Properties of simple tritons ......................... 130
B.2 Binding energy and rms matter radius of a triton ............. 132
Images in this dissertation are presented in color.
viii
Chapter 1
Introduction
Atomic nuclei represent self-bound ensembles of strongly interacting fermions. Experi-
mental and theoretical explorations of the chart of nuclei have revealed many intriguing
features of nuclear matter. Among them, a structural hallmark—the nuclear halo—has
been found in the realm of light nuclei near the limits of particle stability.
In general, the halo phenomenon is a threshold effect occurring in loosely bound
systems, in which particles are held in short-range potential wells. In favorable circum—
stances, a barely trapped particle or particles (or a cluster of particles) may tunnel out
into the classically forbidden region. This “leakage” populates very dilute and fragile
structures near particle emission thresholds. The more loosely the halo particles are con-
fined, the more clearly “the halo stratosphere” is developed.
Besides nuclear physics, halo systems are known or expected to exist in other branches
of physics as well. One of the most extended halo systems known to exist is the atomic
helium dimer 4H82 which is about ten times larger than a typical diatomic molecule and
is bound by only about 10‘7 eV [1]. Halo states have been predicted or experimentally
observed for a range of other systems, such as 3He—3He-39K [2], positron-atom complexes
[3], hyper-nuclei such as 31H [4] among others. A comprehensive review of halo systems
can be found for example in [5].
a ii
A“
1 1.1 H
The gilt-111131
Ilse) PINK"
energies it it I
(‘flllSlllt‘fatl ulx
furn'mtiun Hi-
In a first
of The swirl:
nurlei ("an it.
orbiting it Ii]:
1n quantum l‘
l. the prui
llll‘lllllt.‘
3. and Illi'
a gin-n ;
1t has 1mm in”
it Illif‘ "‘llt‘I‘:
small. in
011110 m.
b' 111" llétli )
(7 and 1]“, 1
t() (1(l\-( l]( l
1.1 Halo in nuclei
The quantum-mechanical tunneling present in halo nuclei produces unexpected effects.
The energy needed to remove halo nucleons is drastically less than particle separation
energies for typical nuclei. Nuclear radii are enhanced; matter and charge radii may differ
considerably. There is evidence that few-body effects may become crucial, leading to the
formation of cluster structures beyond the reach of mean field theories.
In a first approximation, the spatial separation of particles in the halo from the rest
of the system justifies a simplified description with only a few active constituents. Halo
nuclei can be thought of in terms of a few (typically one or two) single halo nucleons
1
orbiting a tightly bound core, thus implying a major role of single-particle properties.
In quantitative terms, it has been assessed [6,7] that for a quantum halo to develop,
1. the probability to find halo particles in the forbidden region beyond the classical
turning point should be more than 50%,
2. and the core-halo configuration should occur with more than a 50% probability in
a given system.
It has been argued [8,9] that for a nucleus to meet these criteria:
a. the energy needed to separate the halo part from the rest of the nucleus should be
small, more precisely less than about 2 MeV 14—2/3, with A being the mass number
of the nucleus,
b. the halo nucleons should occupy s— or p—angular momentum orbits around the core,
c. and the proton number of the nucleus should not exceed ten or so for a proton halo
to develop.
For three-body halo states containing two loosely bound nucleons, the condition b. should
be supplemented by a requirement of hyperlmomentum2 K = 0 or 1. The formation of
1Here, we do not consider less straightforward cluster divisions with tightly bound subgroups of
nucleons, such as 9Be consisting of two 0: clusters glued together by a neutron.
2To be introduced in Chapter 2.
..al
A
figure 11:
scum-s. In I.
[3 explaini‘li
a chargul 1m
nuclei in Fl:
From Ilu'
neutron-rub
“30 l: “'13.
13(11: 13C 4
”iii g: 3m
r".
trill-3 tiff (‘XIN II
da-ay 1i
'Ilqill.
of ordinary. ti-
In IIIIt'lt'ur
mil-1‘2 WIT 11 m
I
.1 ' f I
\ J '
Ll -\ 1.1 +
“1'2," till the at
1136145
.1 thanks Ii
Mitten halo I!"
r,.
[H .
”I (“miplun
y
Proton Number
Neutron Number
p
I pln - Borromean ° l [j pln - Halo
Figure 1.1: Lower part of the chart of nuclei. Stable nuclei are represented by black
“ ,7
squares. In this figure, p stands for a proton, “n” for a neutron. The term ”Borromean”
is explained in Section 1.2.
a charged halo is hindered by the Coulomb barrier. These conditions naturally favor light
nuclei in Figure 1.1 to populate halo states.
From the modern perspective, the best established nuclear halos live among light
neutron-rich nuclei. Examples of one-neutron halo nuclei include the ground states of
11Be (2 10Be + n) [10] and 19C (= 18C + n) [11], excited states in 12B (= 11B + n) and
13C (= 12C + n) [12] and several possible candidates, such as 31Ne (= 30Ne + n) and
40A1 (= 39A1 + n) [5]. In one-neutron halos, the tail of the relative core-n wavefunction
falls off exponentially with the distance between the core and the extra neutron. The
decay length, determined by the neutron separation energy, is typically 4—5 times that
of ordinary, tightly bound nuclei [5].
In nuclear physics, the most obvious three-body halo candidates are light drip-line
nuclei with two neutrons encircling a core. Among them, 6He (2 4He + n + n) and
11Li (= 9Li + n + n) are stereotypical prototypes of nuclear halo systems [13], and they
enjoy all the attention of the present work. 11Li is considered the prima donna of all halo
nuclei thanks to its very small two-neutron separation energy 378 keV [14]. Other two-
neutron halo nuclei include 14Be (2 12Be + n + n) [15], possibly 22C (= 20C + n + n)
[16], and other candidates [5].
For completeness, we should mention other nuclei in which some sort of halo may
3
8351'
be (lt’X't‘lifl'f
]i]((l]\' m 1H
deuti’rH“ 1‘
enerEV {7
states 17
belimnl tn
pui_‘>p111‘dilt in i
bwn sewn II
states. Iii-i};
As an exam
Illt’illlilll-llut.‘
“uric ('Iilit‘ll
exist at lllt" i
in lllt’ [in
particular (it:
1.2 T“
Apart in im
nuclei mm
b
111
(illntl. awn t
Bill'tffllli’itll is
a way that If
example. in.- l
.- ,1 -
“(511- TlllS (K
1r1~
be deveIOped. In the deuteron, for example, the proton (p) and the neutron are very
likely to be found outside the range of the strong interaction. The binding energy of the
deuteron (-2.2 MeV) is in absolute value small compared to a typical nucleon separation
energy (7—8 MeV), arguably making the deuteron the forerunner of all nuclear halo
states [17]. On the neutron-rich side of the chart of nuclei, 8He contains four neutrons
believed to form a neutron skin around the 4He core [18]. On the proton-rich side, the
population of halo nuclei is decimated by the Coulomb barrier. Hints of a proton halo have
been seen in 8B (= 7Be + p) [19], 17'Ne (2 15O + p + p) [20], and some other nuclear
states. Reference [5] contains a more complete list of possible halo states in light nuclei.
As an example of theoretical studies on the existence of halo effects in heavier nuclei,
medium-mass even-even nuclei have been scrutinized in [21, 22]. The authors of these
works concluded that on the large scale the halo phenomenon is very rare and can only
exist at the very limit of neutron stability.
In the present work, however, we shall focus only on light two-neutron halo nuclei, in
particular on 6He and 11Li.
1.2 Two-neutron halo nuclei: 6He and 11Li
Apart from possessing all of the peculiar halo features, the known two-neutron halo
nuclei including 6He and 11Li are Borromean, meaning that the system core + n + n is
bound, even though the binary subsystems core + n and n + n are unbound. The term
Borromean is adopted after a heraldic symbol of three rings which are joined in such
a way that if any one is broken, all three become free [13]. In the helium chain, for
example, 4He binds two extra neutrons, but not one, and the di-neutron is unbound as
well. This odd-even staggering is merely a consequence of nucleon-nucleon correlations.
One then deals meticulously with two correlated neutrons revolving around a core in the
low density regime. Thus, these nuclei are ideal playgrounds to study neutron correlations
in an almost proton-free environment. It is possible that these nuclei give rise to the so-
called Efimov states [23,24].
p 3'
AQ‘
c"_‘ 777‘
+ V__-"
The BU:
cut? and {h
[0:102:03 1“
V 1 7
into 8 C" ”‘5‘ ‘1
distant)?" “'5
Iillln- [111‘ V'Il‘
the two will
These IIEIIIIII'
aiwtmt l 3 I?!
dramatic in 1
Ont" tili‘tt £111
astmptntir [1
DUI? in ii
Single l)iitllltl
In 1" ‘
..lt‘ Of Jill' Is.
Dc» studiml. I
dll'dll'Inly (if [j
p- ,- ,. '7
.Mitdttuiis iii
’19
L1 ‘
“I‘M rPW.‘
81mm ] ' -' ‘
ll-‘lurli‘
L
sit «2 .r '
The Borromean nature of 6He and 11Li implies that, even at large distances, the
core and the valence particles are correlated with no bound binary admixtures. Asymp—
totically, the wavefunction vanishes exponentially with a decay rate depending on the
three-body binding energy, i.e. on the amount of energy needed to break the nucleus up
into a core and two free neutrons. The inverse of the decay rate gives a typical “three-body
distance” within the nucleus, which is about 7.5 fm in 11Li. For better visual apprecia-
tion, this value corresponds to a di-neutron at distance about 6 fm from a 9Li core or to
the two neutrons being on opposite sides of the core at mutual distance of about 11 fm.
These numbers are to be compared with the range of the nucleon-nucleon interaction of
about 1—2 fin and also with the 2.32 fm radius of the 9Li core [13]. The situation is less
dramatic in 6He due to its larger two-neutron separation energy3 of about 970 keV [25].
One then anticipates that many properties of 6He and 11Li will depend chiefly on the
asymptotic part of the wavefunction.
Due to the proximity of particle emission thresholds, 6He and 11Li support only a
single bound state, the ground state. Moreover, these nuclei are short-lived; the half-
life of 6He is 806.7 ms [25] and that of 11Li is even shorter at about 8.8 ms [26]. To
be studied, these nuclei have to be produced artificially. Most information about the
anatomy of nuclear halos has been obtained in reaction processes leading to continuum
excitations and ultimately to the destruction of the investigated nuclei. It is useful to put
the most rewarding experimental methods into their historical context. In the following
short historical overview, we focus mainly on “Li, but some of the experiments have
been carried out for other halo nuclei including 6He.
1.3 Overview of 6He and 11Li: experiments
The history of two-neutron halo nuclei started with the discovery of 6He back in the
19308 [27]. It took three more decades to produce 11Li for the first time [28]. Current
3In what follows, the two-neutron separation energy is taken as an absolute value of the three-body
binding energy, and the two terms will be used interchangeably.
fifi\a I
intents! in I
minus. I:
with otiliIM
were film“ i
(iinsistinz U
by a llli‘dhlll
Consistvnt \'
certainty pr:
mmliél alw s
Soon after. I
high :33: and
('ildrgi"-i‘xi '11. I
implying t1..~t
‘lf‘lliltf‘ tlw II
that only :1. I
lflti‘f'dt’tiuti (-1
Scattered Mm
smaller by a],
.3331.“ of tlti‘ (I]
[11 paramuiq.
[16051112 .341”
”2‘10 S[flu-rim
inn; 9L1 am 21
251136: exrmn r1
:ppf’n‘d in ".11 l
95.30111311an in
listening up a I
The (ta
it. .43
1
TI.) Tr
.—‘
..JJ. Tilt“;
interest in nuclear halos, however, was sparked by the advent of modern radioactive beam
facilities. In 1985, the interaction cross section of helium and lithium isotopes colliding
with ordinary nuclear targets was measured [29,30]. The surprisingly large values for 11Li
were soon interpreted as a consequence of extended neutron densities, a neutron halo,
consisting of a di-neutron coupled to a 9Li core [31]. This speculation was later supported
by a measurement of the momentum distribution of 9Li after the break-up of 11Li [32].
Consistent with the di—neutron model, large spatial extent of the halo was, through the un-
certainty principle, reflected by narrow relative momentum distributions. The di-neutron
model also suggested large two-neutron removal cross sections via Coulomb dissociation.
Soon after, the cross sections of electromagnetic dissociation of 11Li on high-Z targets at
high [33] and low beam energies [34] were found to reach anomalously large values. Later,
charge-exchange cross sections of 829211Li were measured to be about the same [35], thus
implying that the 9Li core is little disturbed in 11Li. One of the first attempts to indirectly
deduce the neutron density profile of 11Li can be found in [36]. The authors concluded
that only density distributions with very long tails consistently reproduce the observed
interaction cross-sections. Furthermore, the angular distributions of 9Li and 11Li nuclei
scattered elastically from protons are similar, but the elastic scattering cross-section is
smaller by about a factor of two for 11Li [37]. In data analysis, both real and imaginary
parts of the optical potentials had to be changed considerably for 11Li compared to global
fit parameters, in order to account for break-up due to the extended tail of the neutron
density. fl-decay represents an interesting alternative for extracting information about
halo structure. Several theoretical works [38,39] have investigated the B-decay of 11Li
into 9Li and a deuteron (d). They concluded that the fl-decay matrix elements are to a
large extent determined by the halo part in 11Li. Experimental efforts in this direction
reported in [40,41] and more recently in [42] provide evidence that the fi—decay takes place
essentially in the halo of 11Li, and that it proceeds mainly to the 9Li + d continuum,
opening up a new means to study the halo phenomenon in 11Li.
The early reaction experiments were extended in later years, see for example reviews
in [43—45]. They include transfer, stripping and break-up reaction studies providing differ-
6
...=.Il
git
ential. mill
att‘nttipiittii
separation f
for 11Li. 1111'
311i] i'iH'tI'lt'
experiment s
111 5])lit‘l
1135 fliil lir‘t'l
flt’lllft'ilih' ('I it
valence iimir
\‘i‘tlmt'ti part
It.) a Strung
[1.3th Tit.)
quest :31 3;]
Experimi
bacnw st, (1,
“J the (“built
b9 turnip] 4“.
nuclei as mh
Strut-um. m“
[’liliis Work.
exotic- IIII('1i -i
ential, rather than integrated cross-sections. Reaction and decay experiments have been
accompanied by precise measurements of static properties: measurements of two-neutron
separation energy by methods of radio-frequency spectrometry [14] and Penning trap [46]
for 11Li, nuclear charge radius determined by laser spectroscopy for 6He [47] and 11Li [48],
and electric quadrupole and magnetic moments of 11Li from nuclear magnetic resonance
experiments [49].
In spite of all the experimental efforts, the detailed structure of the two—neutron halo
has not been deciphered yet. The consensus seems to be that, in 6He, the two maverick
neutrons coexist anywhere between two extreme configurations [13]: a di—neutron with
valence neutrons closely spatially correlated, and a cigar configuration in which the two
valence particles are on opposite sides of the core. In 11Li, the situation is less clear due
to a strong competition between s— and p—waves in the halo part of the wavefunction
[13,50]. The question of clustering in 6He and 11Li is the subject of ongoing experimental
quest [51—53].
Experimental data concerning two-neutron halos collected over the last decades has
become so detailed that theoretical models must be more than merely qualitative to rise
to the challenge. Even simple properties, such as the size of the nucleus, turn out to
be model dependent and are not real experimental observables [54]. The study of halo
nuclei as unstable species via reaction experiments involves tightly intertwined aspects of
structure and reaction physics. Details of the reaction component are beyond the scope
of this work. Nevertheless, reviews of reaction models used to probe the structure of light
exotic nuclei can be found in [55,56].
1.4 Overview of 6He and 11Li: structure theory
Traditionally, theoretical considerations of structure and reactions of halo nuclei have
been dominated by few-body models. Few-body structure models of two-neutron halo
nuclei have built their success around the fact that, when viewed at a distance, the halo
particles are decoupled from the core. Under such an approximation, the core’s degrees
rival
. .).
[figure 1"
nirit’li'is. a II‘
Offrofifiltllll (‘
part. The I:
tt‘vfi'i‘illiir 1W
mirn'm‘rvtiii'
The mrl‘
St'llf"!llitll(' in
Sophistimtwl
his. mule-i
the three-1): If
deer apprt my
method (In t!
the (‘1Il.\tt'r-( ,;
reported in '1
rewarded t hi"
With lill'fi
many~1ior1v p
STUII‘TUI'P 111‘“
I i
nit-tn t0 empl
1' 2 ’
l1.l(IlOIl Mann
1-, .
TM: mode] ”'1
'11‘11 r
-- ... .norli
+1 :C‘fi
Infidel wit it all
O O
:2
O O
microscopic microscopic cluster three-body
model model model
Figure 1.2: Schematic representation of structure models of 6He. In microscopic cluster
models, a microscopically described 4He core is formed explicitly.
of freedom can be reduced, and the wavefunction factorizes into the core and the valence
part. The many-body problem then reduces to a three-body one—core + n + n———held
together by effective core-n and n-n interactions. For 6He, the transition from a fully
microscopic to a few-body picture is schematically depicted in Figure 1.2.
The early di-neutron models of 6He and 11Li, such as [31], turned out to be too
schematic to quantitatively describe experimental data and were soon followed by more
sophisticated three-body approaches. In the first generation, the three-body models of
these nuclei treated the core as a completely inert object. Several methods of tackling
the three-body problem were applied, mostly to 6He and 11Li. They include the Fad-
deev approach [13,57,58], the hyper-spherical harmonics method [13,59], the variational
method on a harmonic oscillator basis [60], the two-body Green’s function [61], and
the cluster-orbital shell model [62,63]. Some calculations within a pairing model were
reported in [64]. In all their generosity, the three-body models of the next generation
rewarded the core with some degrees of freedom, namely with rotational modes [65].
With increasing computational power in recent years and new techniques to solve
many-body problems, ab-initio microscopic competitors have emerged in the field of
structure models of light exotic nuclei. The microscopic nature of these models allows
them to employ realistic nucleon-nucleon and three-nucleon interactions. The Green’s
function Monte Carlo model has been successfully applied to light nuclei up to 12C [66,67].
The model reproduced the three-body binding energy and radius of 6He. The no—core
shell model [68] is another sophisticated approach, which as its name suggests, is a shell
model with all particles active in harmonic oscillator shells; i.e. there is no inert core
Evil
Dim
hit“, in statii
and “Ll 7
dt’lldlzllli‘ 1"
The) built 15
Their Hl’i’ii‘
.\11iit‘;t(’ (‘iifil
dynamics mt
Stilllt’W119'
models. in \i'
maitcls. This
internal strut
tun-s can aim
Stut‘hdstit' m:
helium [76: ill
strength. the i
and ”Li. (m.
.111 these 111m]
miitii‘in,
Our nt'en'ii
“Eddy f1i.\])i-r:
Ullf‘it‘iins ll] flit
Wan fields. p,“
appl‘lpriate. a.
.11 s’irtm'er. t 1
1i“ 1;
mean fielil (If 11
B, .
irrommn nut-1
1:.»-
Mill field cairn
(trifle
Vumtion ”11'
it
I a good (luv-
Tl
like in standard shell model calculations. The model has been applied to both 6He [69]
and 11Li [70]. The fermionic molecular dynamics and the antisymmetrized molecular
dynamics represent conceptually similar approaches to the problem of light nuclei [71].
They both use superpositions of Gaussian wave packets for single—particle wavefunctions.
Their application to helium isotopes can be found in [72,73]. As in Green’s function
Monte Carlo, the structure of 11Li has not yet been successfully described by molecular
dynamics models.
Somewhere between few-body and truly microscopic models are microscopic cluster
models, in which some degrees of freedom are frozen to reduce the computational de-
mands. This is achieved through the formation of microscopic clusters with a simplified
internal structure within the nucleus being modeled. To a certain extent, cluster struc—
tures can also be recognized in some of the micrOSCOpic models mentioned above. The
stochastic variational model [74] and its multi-cluster version [75] has been applied to
helium [76] and lithium [77] isotopes. With simpler phenomenological forces of adjusted
strength, the model has been able to reproduce basic (three-body—like) properties of 6He
and 11Li. Other examples of microscopic cluster models applied to 6He include [78—80].
All these models rely on Gaussians of one sort or another to describe the inter-cluster
motion.
Our overview of structure models would not be complete without mean field theories.
Widely dispersed halo particles barely feel the short-range nuclear forces exerted by
nucleons in the core. As a consequence, valence and core particles experience different
mean fields. For Borromean systems in particular, the term mean field is probably not
appropriate, as the correlations between halo particles are crucial for the overall binding.
Moreover, the last neutron in the core + n + 11 system can not be bound in the localized
mean field of the core + n subsystem since such a bound subsystem does not exist in
Borromean nuclei. The importance of unusually small neutron separation energies for
mean field calculations was recognized early on [81]. In later shell model calculations,
configuration mixing and adjustments to residual interactions have become unavoidable
for a good description of exotic nuclei towards the drip-lines [82,83]. In general, mean
9
pail
field Alli—m W
1.5 m.
The uhviulh
exact ”Hill“
the relatin- ll
ptupttlt. tiiv'j
1913111305. ()3;
11:11 temple ‘in
ing t-nizipiimt:
rtsults fur in.
flit.iili.'1>' will \I.
We are 11‘
mp9 with tit-i
mentioned str
{intents trmlir
T110 ("rm-j“
dUuhle-(«lgpd q
litittt‘C‘t‘n (1 m. ,1
Cation Of 1]“, II
“13351 that (*u
1”“ 0n the
h
()I 1“
I110 mlffilf‘] r
.18.
&
ram . .
attic halo It
is al
a. p, .
Ilin
tag from the
Ra‘s O f )y .u
(
A
field approaches have found it rather challenging to obtain a reasonable description of
halo effects in light nuclei.
1.5 Motivation for present work
The obvious advantage of few-body structure models of two-neutron halo nuclei is the
exact treatment of halo dynamics. These models provide clear, intuitive insight into
the relative motion between the core and valence particles, and as long as implemented
properly, they are well suited to capture the long-distance halo characteristics and cor-
relations. On the other hand, micrOSCOpic models tackle the many body problem in its
full complexity. Thanks to our advancing knowledge of nuclear interactions and increas-
ing computational power, brute force ab-initio models now yield very accurate structure
results for many light nuclei. It is reasonable to believe that, sooner or later, ab—initio
models will succeed in producing an accurate description of halo nuclei.
We are now in the position to ask why we need yet another structure model to
cope with two-neutron halo nuclei. The answer is buried in drawbacks of the above-
mentioned structure models and the lack of connection of some of them to reaction
theories traditionally formulated in a few-body framework.
The crucial assumption of few-body models, a macroscopic core, turns out to be a
double-edged sword: on one hand, it allows us to focus on the most important correlations
between core and halo nucleons, on the other hand, it is undoubtedly a (crude) simplifi-
cation of the many-body problem. To argue in favor of inert cores, some authors indeed
suggest that core polarization in halo nuclei is suppressed compared to normal nuclei [84];
but on the other side, there are works that admit the possibility of less inert cores inside
halo nuclei [48,76]. Despite the occasional strong claims by few-body practitioners [85],
realistic halo nuclei are unfortunately not ideal halo systems; the simple halo picture
is always obscured by small idiosyncrasies, and one has always to check that the core
is really unperturbed to justify the simplified inert-core few-body approach [86]. Stem-
ming from the simplified picture of the core, probably the two most severe drawbacks
10
.ail
Qt
U
A
"‘2' :n": ‘
of few-iv "1‘
inwramiuti‘
amid-nun“?
Prm'iiii" (iiifi
halt) 1”m ”1'
rpprntlut‘i‘ *4
of the win '1"
fort-cs. Fur? ii
intuitions 1’“
missing in ii;
Lard in hit M
Sumo of I
gr-ttipit‘ 1(‘1‘:I>l
The Hlli'fll.“
use phenom
is wrijmg at
The first out
unrlvrstanriii
picture. our
the full llllt'
E‘v'f’ll lilting]
Shin Wit [“1
PTOYirliIig M
Complt’ldl it ,1
adf-‘qliar-y l’)‘.
PITJTIL I}
I: 2‘ "
mite (alt-til
b.
‘5‘" ' '
‘3- Clit‘.‘
is
I. ‘—
AIIJXI‘T'.\.‘
of few-body models are the lack of exact antisymmetrization and the usage of effective
interactions [87]. Several Pauli blocking techniques have been developed to account for
antisymmetrization in few—body models, however, when compared side by side, they may
provide different results [88]. Effective interactions, especially those between the core and
halo particles, are not necessarily known. Normally, the core-n potentials are adjusted to
reproduce some set of experimental core-n findings and the three-body binding energy
of the whole nucleus, or attempts are made to derive them from the underlying nuclear
forces. Furthermore, there are indications that for reaction calculations three—body wave—
functions perhaps require additional renormalization to account for microscopic effects
missing in the inert-core approximation [89]. Nevertheless, few—body models are presently
used in most reaction calculations involving halo nuclei.
Some of the above-mentioned drawbacks of few-body models are eliminated in micro-
scopic (cluster) models with halo particles made indistinguishable from those in the core.
The microscopic treatment allows one to antisymmetrize wavefunctions properly and
use phenomenological or realistic nucleon-nucleon (and three-nucleon) forces. So, what
is wrong with microscopic (cluster) models? Well, one could object to several things.
The first one is the missing connection to reaction theories, a link so important for the
understanding of halo species. To feed reaction calculations formulated in a few-body
picture, one would have to extract the necessary information about halo particles from
the full microsc0pic wavefunction, a task that is by no means trivial computationally.
Even though recently we have witnessed some progress in this direction for two-body-like
(but not halo) projectiles [90,91], most microscopic structure theories are still far from
providing such few-body-like information relevant for three-body-like halo nuclei. This
computational obstacle is accompanied by a more fundamental physics question of the
adequacy of microscopic models in the asymptotic regions.
Horn the previous short review of structure theories it has become obvious that to
make calculations feasible microscopic (cluster) models exploit computationally tractable
bases. Chief among them are the Gaussians and harmonic oscillators. One must remem-
ber, however, that at large distances, where the halo nucleons are almost liberated from
11
—
1'11
Qt
1119 (1UP. 1
capture th
as argued
itrds hint?
tit'itis their:
retreat 11211
are in gritt-
1111('T'I>('Ii]lil.
of ('UIlW‘TEII"
antw this ('
Ravehinrth,
1391,5011 it;
crt;rsttipir q;
“1‘11 to mix
l‘v‘t'i'I-nplnr‘ Ill
ITllST‘dIIi‘i" Hui
CUH‘II31 (if it
Wat't‘flillifiitm
halt) lllli'll’l tit
1121108).
In 111(‘11. .
Pff)tll.l(‘l (if d n
0 ~ . .
txhonmrit-t‘
r I
it " '
tsptttt of lilt‘ ]
m;- ‘
it): flllf‘lf‘UIl‘
3M1? .
1 I '
1 1’15; ()f “(1
the ..
mdW’fllIirt [I
the core, the wavefunction falls off exponentially. In principle, it should be possible to
capture the slower exponential decay by using a large Gaussian or oscillator basis, but
as argued in [5], quality precedes quantity in the halo world; that is the actual shape of
basis functions matters more than the size of the basis. In other words, the basis func-
tions themselves ought to possess the correct long-distance functional form to produce
correct halo asymptotics. For this reason, the authors of [5] concluded that Gaussians
are in general not at all suited as a computational basis for halo nuclei. Moreover, most
microscopic calculations are variational with the binding energy used to assess the rate
of convergence. In general, the convergence of the total binding energy does not guar-
antee the convergence of other observables and definitely not the convergence of the
wavefunction in asymptotic regions.
Based on the arguments presented, one can conclude that both few-body and mi-
croscopic structure models have their appealing aspects as well as their drawbacks. We
wish to mix the best of the two approaches to create a microscopic structure model of
two-neutron halo nuclei that would describe simultaneously short- and especially long-
distance regions and allow us to link .the structure and reactions of these nuclei. The
concept of a microscopic cluster model with a carefully chosen functional form for the
wavefunction seems to be ideal to meet our goals. Hereafter, the model of two-neutron
halo nuclei developed in the present work shall be referred to as MiCH (microscopic core
halos).
In MiCH, a two—neutron halo nucleus will be described by a properly antisymmetrized
product of a microscopic core and the valence part consisting of two individual neutrons,
or schematically II! = Acore‘val(core x valence). We shall use terms “core” and “valence”
in spite of the presence of the core-valence antisymmetrizer Acme-"al which, in principle,
makes nucleons from the two parts of the wavefunction indistinguishable. A more precise
meaning of “core” and “valence” will be provided as we go along. At large distances,
the wavefunction naturally decouples into the three-body—like form \I! ——+ core x n x n,
whereas at short distances it is equivalent to a fully antisymmetrized, many-body treat-
ment. To bind the nucleus, effective nucleon-nucleon interactions shall be employed. The
12
P-al
lllf‘lfify (1i’\‘r‘
nuclei. The
ttt‘t.“i-Iii;*ltt rt iii
in one step
prams is iii
112110 pit ’11“
1.6 0‘
1n the prwr
111-3616 Hi the
haptvr '2 it”
PHIPIIIS of it}
10ml for 111i.“
118 nt‘utrtitis
6H9 and 111.1
mtiflt‘l that lit
Chapter 4 by
1119 COII‘iputgn
tron of vari
B
ill it
2151c strtlt't lit
a ..
re (‘Olllpitl‘t‘ll
.11th.\$1(ill. tht
S’fl - '
t .irture input
outlook in f '1
t.-
theory developed in this work is designed to cope with bound states of two-neutron halo
nuclei. The link between structure and reactions will be established for simultaneous
two-neutron transfer. In this reaction channel, the two valence neutrons are transferred
in one step from a halo projectile to a target nucleus. The transition probability of this
process is directly proportional to the overlap integral between the original two-neutron
halo projectile and its own core.
1.6 Outline
In the present work, we elaborate on all aspects of MiCH. First, the two major building
blocks of the wavefunctions—the valence part and the core—are discussed separately.
Chapter 2 focuses on the valence part. A particular three-body model is described, el-
ements of which are later incorporated into MiCH. To find the appropriate functional
form for the valence part, the three—body dynamics between the core and the two ex-
tra neutrons is studied in interaction-free regions. Chapter 2 also contains results for
6H6 and 11Li studied within a three-body approach. Chapter 3 presents a microscopic
model that meets requirements imposed on the core. MiCH is then finally assembled in
Chapter 4 by putting the core and the valence part together. That chapter also includes
the computational background needed for evaluation of matrix elements and optimiza-
tion of variational parameters. Chapter 5 contains results for 6He studied within MiCH.
Basic structural features of 6He are elaborated on, and the results obtained within MiCH
are compared to those from other models and to experimental data. As part of the
discussion, the two-neutron transfer reaction p(6He,4He)t is studied using microscopic
structure input for 6He modelled in MiCH. The work finishes with the conclusions and
outlook in Chapter 6.
13
...;I
Chap1
Valen'
Con: retry t«. l s
bound and t 1
serrahlt‘s. Atty
at‘t‘tjitiiit the: 111:
(iii? Sllltjt‘l'l in
tiUC‘li.1.tlt iIltt‘I'm
to pin tit with it:
added Bum )Iiii
any other my“
with the few-1n
the Wat'efunt't it
CUT" 211111 a lllhw
neutrons rt-lut it.
111 1110 r-m‘rt
Established thri-
model will 1w r
PM thrtre ht it 1 V
mode . ingrmlit 'ti
Three-1 ‘
.. ind)“ Intuit.
Chapter 2
Valence part
Contrary to standard nuclei, valence particles in two-neutron halo nuclei are weakly
bound and the tail of the wavefunction offers large contributions to most physical ob—
servables. Any structure model aimed at the description of halo species should take into
account the fact that the loosely bound neutrons swim in distant, low-energy regions and
are subject to an interaction which is closer to the free rather than in—medium nucleon-
nucleon interaction. Thus, a proper treatment of the asymptotic regions is vital if one is
to pin down any observable sensitive to the spatial extent of the nucleus. Moreover, the
added Borromean peculiarity of two-neutron halos implies pure three-body rather than
any other asymptotics. Few-body models are especially well suited to cope successfully
with the few-body dynamics and asymptotics of two-neutron halo nuclei. In Chapter 4,
the wavefunction in MiCH will be cast as an antisymmetrized product of a microscopic
core and a three-body-like valence part describing the relative motion of the two valence
neutrons relative to the core.
In the current chapter, we focus on the valence part. To do so, we outline a well
established three—body model [65,92,93]. To avoid repetition, the mentioned three-body
model will be referred to as “the three-body model”. First, we introduce coordinates
and three-body basis sets used to attach the halo neutrons to the core in the three-body
model, ingredients to be incorporated later into MiCH. Then, we outline details of the
three-body model beyond what will be built into MiCH, such as interactions, the Pauli
14
..-;I
principle. anti
will have no it
than out in it] i:
are inclnrlml it:
plii‘iiishi‘ti lit r:
in :11 1: an: reign
flt‘t’l'lt‘fl I‘lr 1“ '1:
2.1 Coc
The key inqrm
spherical ft irnm
Dh‘CSits. W115 1m
rial tit in the a try
tii'iurtnittit Ill 1111
hyper-radial 1n
Laznt'rre hyper
for (“lam y. i
chapter. "\‘ali‘nt
the Pmpcrt it *.\ (
, . .
fUll informat it ill
the meaning of
T0 50!? ht iv:
tare ‘9‘ n + n b
TUlf . l
1] 11.“(1dr(‘ \‘)]
wt: Spurious mm
”11*? at
d “Puff”!
N
)1] . ‘
g fringe (. (if.
principle, and the actual way of solving the three-body problem. In later chapters, we
will have no use of these extra aspects of the three-body problem, but it is useful to lay
them out before us to perform three-body calculations for 6He and 11Li, results of which
are included in this chapter. The results of three-body calculations for 11Li were recently
published by the author and collaborators [50]. For 6He, calculations originally published
in [94] are repeated to reach results that were not included in that article but that are
needed for comparison with results obtained within MiCH for this nucleus.
2.1 Coordinates and bases
The key ingredient of the three-body model is the Schr6dinger equation in the hyper-
spherical formalism. The hyper-spherical method, which had been used in other areas of
physics, was brought into nuclear physics in [95] with the aim to develop a general nuclear
reaction theory. The value added to three-body models in [65,92] was the introduction of
deformation and rotational degrees of freedom to an otherwise inert core. The Sturmian
hyper-radial basis exploited in [65, 92] was later in [93] replaced by a more suitable
Laguerre hyper-radial basis [96].
For clarity, we should define terms “core” and “valence” more precisely. In the current
chapter, “valence” will refer to all features of the three-body core + n + n system except
the properties of the core, i.e. it will encompass spins of the two neutrons as well as the
full information about the relative motion between the three bodies. Later, in Chapter 4,
the meaning of these terms will be elaborated.
To see how the three-body model is assembled, let us first analyze a three-body
core + n + n bound problem in interaction-free asymptotic regions1 where, as argued
in [5], one ought to employ a basis with appropriate exponentially decaying form. At this
point, we are solely interested in relative motion between the three bodies. To eliminate
the spurious motion of the total center of mass, only relative Jacobi coordinates between
core and neutrons are used as shown in Figure 2.1. In principle, the two sets of Jacobi
1Long range Coulomb effects are absent due to charge neutrality of valence particles.
15
..-:l
(tit‘itt‘litlati's
cast in any ill
rttx‘irtlinatts l5
Partit‘lp'llkp 1)
Width in 1m 1 a:
\
Wllll lightly <
1! is WNW“
par1511li‘l'llll]litt
it"‘r' '
.19“? 1]- ari‘ at)
tall ‘dllll they '
momenta 1: and
particle t-oort ll 11'
IL"; ,
an advantage
they allow the t
rat '~
lidl equation.
1
.n . ~
,ptrsphcrit‘al r
I 31 '
iii-trig-
coordinates—~Y and T—are completely equivalent and the three-body problem can be
cast in any of them. The main advantage of Jacobi coordinates over other sets of relative
coordinates is that the operator of kinetic energy decouples into two independent single-
particle—like pieces with no cross term:
it? 1 1 h?
T=——- —A~ +—A,~,2 =
2m #1 1.1 #2 [A5 + A37] , (2.1)
7275
where m is the mass of a nucleon. Then, the interaction—free three-body Schréidinger
equation becomes:
T12 (5,37) = E3body¢ (13,37), (2-2)
with E3body < 0 being the three-body binding energy.
It is convenient to seek the solution of Eq. (2.2) in the form with angular and radial
parts decoupled, schematically:
12(59') 2 HOT, 3»le (931223,,(931), (23)
where Y; are spherical harmonics (for now, their projection quantum numbers are omit-
ted) and they take care of the angular part of Eq. (2.2). We stress that the orbital
momenta la; and ly are associated with Jacobi coordinates, rather than any sort of single-
particle coordinates. Next, hyper-spherical coordinates from Figure 2.1 are involved. The
main advantage of hyper-spherical coordinates is that, as it will soon become obvious,
they allow the transformation of the original Eq. (2.2) into a one-dimensional, hyper-
radial equation. The radial function H(:r,y) can be equally well written in terms of
hyper-spherical coordinates p and 0, i.e. ’H($,y) = ’H(p,l9). Plugging (2.3) into (2.2)
yields:
a 1
— (p555) + $19] H (12.6) = EgbodyH (a9), (2.4)
16
/
I
Let us ("HIM
I ohjett: are
ilk? N'l_~ If.
’—
RPIMIW‘ .14“ 1
tilt" la>t J.“ n
t'vrztcr of rm:
‘
H l
l
H:
I.)
l
\
Th“ VUEHHU‘ ('
litre. Q, (., ,m]
II. ‘\(’xt. r.:\(.‘
Wllll (limvnsin
\
\
\‘
NUtp .
0 -\thdl Shh
"I
\ (
.‘ '21 RIM]
and
”If“ ln'p. ‘r-
Let us consider a three-body system core + II} + n2. In the laboratory frame, the three
objects are at positions 77007.8, 7"}, and 7"}, . Then, there are two different—Y- and T-
like—sets of Jacobi coordinates :75 = {in = 1, 2, 3}:
Relative Jacobi coordinates 531 and f2 connect centers of masses of subgroups of objects;
the last Jacobi coordinate 53 (not shown in the graphics) is equal to the position of the
center of mass of the three-body system in the laboratory frame:
Y T
51 = in] — 77core 531: F112 _ 71.721
_. _. (77111 + Acorefowre)/(Acore + 1) 52 = (F712 + Fn1)/2 — Fcore
1172 = n2 —
53 = 77CMS = (Acorefcore + F721 + Fn2)/A
The volume element corresponding to the two relative Jacobi coordinates is:
dV = dildfg = x? 323 dxldxgdflldflg.
Here, (2,- comprises the standard polar and azimuthal spherical angles associated with
5,7. Next, rescaled relative Jacobi vectors are defined as:
5=VH15L .77=\/M2$2
with dimensionless reduced mass factors:
Y T
#1 = Acme/(Acme +1) #1 = 1/2
#2 = (Acore +1)/A #2 = 2Acore/A
Note that spherical angles associated with :i" and 37 are the same as {21 and (22, i.e.
(2x = (21 and fly = {22. Finally, the hyper-spherical coordinates, the hyper-radius p
and the hyper-angle 0, are introduced as:
x=psin0, y=pcos€.
The volume element now becomes:
dV = (p1,u2)_3/2 p5 sin2 6 cos2 6 (1de ML; de.
Figure 2.1: Definition of Jacobi and hyper-spherical coordinates for a three-body system
core + n1 + 112. Acme and A are the mass numbers of the core and of the whole system.
17
.-al
gu
with the gI‘H-
Tlds upwmm
'llifllllt’lll it it.‘
spa-[mm ..f e‘
In 93 flu: v.
For the purpn
me rt'laIiumlA
-1.
as.
_ I
R hvre PHI -
Jtll‘
lElllf‘S of the h‘
For -
d myt,“ I)”
(..-
Sliilf’rlm]
Wlllllll
with the grand-angular operator A2:
2.5
—2—n 26 6_6 86 sin2 6 cos2 6 ( )
This Operator contains dimensionless magnitudes ($0,; + 1) and ly(ly + 1) of orbital
momenta as traces of the orbital motion. The grand-angular operator has a complete
spectrum of eigenfunctions enumerated by hyper-momentum K:
A2
:1
Of dbl'lllpiuilt';
hyper-rm lial l»
I... 5 '
Mere Lu; 4.,
((1!)
A few ('(lllilllt‘ll
this ba~i> is jlh
es liyper-ratlial
contains the «log
respect to [ilt‘ W
whereas at large hyper-radii, the hyper-radial part of the wavefunction falls off as:
p—+oo 5/2
Mp) flHam-Hp), 73(1)) ——+p_ exp(-I€p). (2-15)
Indeed, the three-body binding energy buried in it through Eq. (2.13) determines the
decay rate of the wavefunction. The smaller the binding energy the more pronounced
the halo effects. In the three-body approximation, two-neutron halo nuclei posses only
the long-distance asymptotics in Eq. (2.15) due to the non-existence of any bound binary
subsystem, and because the hyper-radius is invariant under the change of Jacobi sets, the
asymptotics are the same in both Y and T Jacobi sets. Having in mind the importance
of asymptotical behavior of the wavefunction, the three-body model employs a Laguerre
hyper-radial basis introduced in [96]:
1 ”lag! 5 P 1 P
7?,” p,p = —— —————-——L —— exp ——--—— , 2.16
lag ( 0) p3 ("lag + 5)! "lag pO 2 P0 ( )
where Lilagm/po) are associated Laguerre polynomials of the order may = 0, 1,2, . . ..
A few comments regarding this basis are appropriate at this point. First and foremost,
this basis is just a suitable mathematical basis, elements of which can not be interpreted
as hyper-radial eigenfunctions of the physical three-body system. The basis explicitly
contains the desired exponential part. The basis functions Rnlag are orthonormal with
respect to the weight p5 which occurs in the hyper—spherical volume element in Figure 2.1:
00
R R , 5d :5 2.1
[0 nlag(p,po) flag“) po)p p nzagmfag ( 7)
with the help of the relationship 22.2.12 from [97]. Moreover, the basis is complete for any
value of po. This fact is of great importance because any p0 can be used in calculations
and yet the proper asymptotic exponential behavior determined by the a priori unknown
three-body binding energy, or n, can be reconstructed. Last, the hyper-radial basis is
privileged in that its index may is not restricted by quantum numbers attached to the
20
sail
splieriml ant- l
Basfll on
W‘th‘lell'I it in
Wilt‘fP 6 1> (‘1 .
Jim. (lviiim's I
parity is (li‘It‘I‘I
are spins of the
to all but lllt‘
,1 = {Ii-.11. I!
mmlel: later. ii;
particles.
hi this: sow-ti.
nil! . - .
“m- be In (lint)
ingredients of th
2.2 Othc
lti the tlirwulmrli
0:7,, ,
...}.Natt‘s of Ill"
spherical and hyper-angular parts of the wavefunction.
Based on these schematic arguments, a single basis term for the valence part of the
wavefunction is written as:
7.0
J7?
val
Z Rnlag (piPO) W7 1 (Baflxaflananng) (218)
"lag 711a! val 13a
where W is a generalized hyper-harmonic function in an LS—coupled product form:
_ ’ny
gym“JgalwaaxinannlaXng) "" ‘PK (mx
[ [1293“le ‘8 Yly(Qy)] L ‘8) [X111 (8) Xn2ls $119)
val
J30) denotes the total angular momentum and the parity of the valence part. The
parity is determined by orbital momenta 1;; and [3, as 7r = (—1)lx+ly. an and Xng
are spins of the two valence neutrons. Index 7m; comprises quantum numbers related
to all but the hyper-radial part, as well as the Jacobi channel identifier Y or T, i.e.
7w; 2 {K , Ix, ly, L, S, Y/ T}. This form of valence terms is sufficient for the three-body
model; later, in Chapter 4, each valence term will be enriched by isospins of valence
particles.
In this section, we have introduced only those elements of the three-body model that
will be in Chapter 4 incorporated into MiCH. The next section details the remaining
ingredients of the three—body model.
2.2 Other ingredients of the three-body model
In the three-body model, the core is a macroscopic object with states (PJgrorefi) that are
eigenstates of the core’s intrinsic Hamiltonian:
hc°re<€>¢Jam<€> = Eerore‘I’Jarfi- (220)
21
.-al
i
To iziiprm't‘
the core lN'l
of a “drill“
are the rum
gems as ll."
A (111ml ii
in light nutl
riiitor-likv ft‘il
deft irmat it in
the groiiiiil s
‘ Be 'll'ii. \W
‘1 ‘ - .
is not known
reasurialilv pl.
C‘dlt‘tllalliius f,
lll this “'HI'k_
Having rlj.
¢’J‘
(are '
the Ill
iv - 2
JH
NJ
.17
f l
W9:
.th numbers r
l
a. _
st exprt'sxiun
fem" '
15 m the ("x‘
"l‘ "
uchlllIIl leIH
lhf pa
To improve upon the inert-core assumption, the adOpted three-body model assumes that
the core behaves as a macroscopic deformed rotor and includes the lowest energy states
of a rotational band built on the ground state of the core. The eigenstates of the core
are the rotational matrices with Euler angles as coordinates 6. Quadrupole deformation
serves as the collective degree of freedom.
A question may arise whether the assumption of a deformed rotor-like core is justified
in light nuclei. Even though it is true that light nuclei in general do not show genuine
rotor-like features, some of them are known to be deformed. In such cases, the quadrupole
deformation can be adjusted to reproduce the strength of the E2 transition between
the ground state and the first excited state in the core, as in the case of 10Be core in
11Be [10]. When the core does not exhibit rotor-like features or the E2 transition strength
is not known, the quadrupole deformation is a free variational parameter taken from a
reasonable physical interval, as in the case of a 9Li core in 11Li [50]. Besides three-body
calculations for 11Li in Section 2.3, we will not rely on the assumption of a rotor-like core
in this work.
Having the basis in Eq. (2.18) for the valence part and the states of the core given by
<1) Jgo'r e, the three-body decomposition of the total wavefunction is finally written as:
Jgore nlag “Yual Jgal J
with numbers c being linear expansion coefficients. It is understood that parities 7r in the
last expression implicitly carry the same subscripts as corresponding J. The number of
terms in the expansion is controlled through the number of included states of the core,
maximum hyper-momentum K, maximum order may of hyper-radial basis functions, and
the parity requirements 7r 2 more anl and “val = (—1)lx +131. Additional constraints on
the basis may be imposed by limiting the maximum orders lg, lg of partial waves. To
take advantage of the completeness of the hyper-radial basis in Eq. (2.16), the nonlinear
parameter p0 is the same in all terms in Eq. (2.21). Although, in principle, both Y and T
Jacobi sets of coordinates work equally well and could even be mixed in the wavefunction,
22
8.5"
).
f‘fi'rc “2.2: l .
H1I*I‘l€.’l. Ihm
’V'P [Hz-emit” ll
A» ‘
the thrw-ln ulK
far ream ins ex;
ln St‘t'llilll
Cit-'hltli‘nxl, “i-
lWlV l'léilllilll iii
,7.
0‘0". fu‘ti—lnuh
pfnsihh‘i (llrl‘Q‘-ll
H : 7‘14““an [:-
ln flirt-(.Lluuh
those hetwm'ii thi
table! potent lill
.\
p'iélt’nlléils are no!
Uriel. the con-~11
Empirical form wit
hr the core + n s.“
Ito-{l
e in this chap
COI‘G
Figure 2.2: Pair-wise coordinates used to define two-body potentials in the three-body
model. These coordinates are essentially the same as vectors 551 in Figure 2.1, but they
are renamed here for the purpose of potential definitions.
the three-body model starts with the wavefunction written solely in the T Jacobi basis
for reasons explained later in this section.
In Section 2.1, a simplified three-body problem in the interaction-free region was
considered. We now extend our considerations to the full physical space where the three-
body Hamiltonian contains the kinetic energy T, the intrinsic Hamiltonian of the core
hm”, two-body interactions V‘me‘" and V"‘” for all pairs of interacting bodies, and a
possible three—body interaction V3b0dy:
H = T+hc‘”‘e(€)+ere‘”(rcore_n1 , §)+vw"€—"(ch~e_n2,§)+v"-"(Fn,_n2)+v3b0dv
(2.22)
Figure 2.2 depicts the corresponding pair-wise coordinates. The operator of kinetic energy
expressed in Jacobi coordinates appears in Eq. (2.1).
In three—body models in general, the exact form of two-body interactions, especially
those between the core and the valence particles, is rather uncertain. In some works, inter-
cluster potentials were derived from underlying nucleon-nucleon interactions [98]. Such
potentials are non-local, however. In most three-body applications to two-neutron halo
nuclei, the core-n interactions are not founded microscopically; rather they are given an
empirical form with parameters adjusted to reproduce some set of experimental findings
for the core + 11 system and possibly the three-body binding energy. The three-body
model in this chapter adOpts the later approach. The exact form of potentials will be
23
.-ai'
Qt
5b.,
’«X‘l ‘
shown lzitt‘r
For Stirrit‘ (‘0
wring plum:
available. ()1
lying energy
from a sing-'14
Curt" + n 5‘"
for the: inter.
potential l> i;
thrt‘tklu‘uly sx'
up. an athliti.
that Illi" IIIIH'
belJ'llllet'l)‘
lll {Hr—lit ni
Hilt'lmins, I‘ll“.
lff‘dlmi'm ..f tl
undr'r Ii
10 INT”
, .
Dds l.“
i. - '
a} impn‘.
who '
r011 IS tliw
and Tare total '
anti"-
.55mnwtrv m
attihi l)&L\l.\' \k'lii
thetl
. trawl
)mlv
. n
This. -
o d('(‘()llllI
Lent
Mn. thr- 111ml
110.“;
H.
of prm'jdp
.lfflfJS that tfl“
shown later in this chapter when three-body results for 6He and 11Li will be discussed.
For some core-n systems, such as 4He—n, the interaction can be fitted to reproduce scat-
tering phase shifts. For other systems, such as 9Li-n, the scattering phase shifts are not
available. One then adjusts potential parameters to reproduce experimentally known low-
lying energy levels in the core + 11 system asserting that the levels can be constructed
from a single neutron orbiting the core. Unfortunately, low lying energy spectra of the
core + 11 system may not be well known, as it happens to be in the case of 10Li [99].
For the interaction between the valence neutrons some sort of realistic nucleon-nucleon
potential is used. When the two-body interactions alone are not sufiicient to bind the
three—body system by the experimentally observed amount against the three-body break-
up, an additional three-body interaction may be introduced. Overall, it becomes obvious
that the uncertainty due to interactions remains one of the major drawbacks of three-
body models.
In few-body models, it is impossible to account prOperly for the fermionic nature of
nucleons. The wavefunction can not be fully antisymmetrized due to the macroscopic
treatment of the core. The wavefunction can, however, be explicitly made antisymmetric
under the permutation of valence neutrons, a requirement easily achieved in the T Jacobi
basis by imposing:
lg; + S + T 2 odd (2.23)
where lag is the orbital momentum of the relative motion between the two neutrons, and S
and T are total spin and isospin of the valence part. For two neutrons, T = 1. This simple
antisymmetry condition considerably reduces the number of available channels in the T
Jacobi basis when compared to the Y set of coordinates. It is for this very reason that
the three-body model starts with the wavefunction in Eq. (2.21) written in the T Jacobi
basis. To account approximately for the Pauli blocking between the core and each valence
neutron, the model space is restricted further. The three—body Hamiltonian in Eq. (2.22)
does not provide information about the internal structure of the core. The model then
assumes that the fictitious core’s neutrons sit in the same core-n potential well defined
24
s.-.”
l
r]
l
] I
r. . _'1*‘
for \ali‘fl“*’ l
halo systviiis
to be orriipi
Spat? ‘intlliil
{ULHIlIlt‘ll mi
forblllilgjn 51,.
flit" linen
(lidii'nitilizti? i.
lllflt'iliim can
\) lzt‘ll lli’i (‘nr
ti-i'rrlitf'ii‘rits r,
With energy a:
Tl
en .
mg of lllt“ 11
St
mm are pro].
lfOlIlpllf'aIWl f
_ ii!‘
In atrtiial ("I
- :IIIdry I
\' ‘11, (I. Yll
a:
._p . .
illlffflls of I}
' ~ 10
of J,
.dflfjbi .
Sl'fifr
'- 'Hi:
for valence neutrons which, as we have argued in Chapter 1, may not be adequate for
halo systems. Nevertheless, the lowest states produced by the core-n interaction are said
to be occupied by neutrons in the core and as such should be eliminated from the model
space available for valence neutrons. Several different techniques exist to suppress the
forbidden core-n states [88]. In the three-body model described in this chapter, such
forbidden states are projected out before diagonalization [65].
The linear coefficients 0 in Eq. (2.21) can be obtained through the energy matrix
diagonalization. Upon arbitrary reordering of terms in Eq. (2.21), the three-body wave-
function can be schematically written as:
\II = Zo,iII,-. (2.24)
2'
When no core-n states are projected out of the valence model space, the expansion
coefficients c,- are obtained by solving a set of simultaneous linear equations:
Hij cj = EIz-j cj (2.25)
with energy and overlap matrix elements defined as:
Hij = (‘I’ilHl‘I’jla Iij =(‘1’il‘1’jl- (226)
The operator H is the Hamiltonian from Eq. (2.22). In Eq. (2.25), E is the binding
energy of the nucleus. In the three-body model, E = E3body- When the forbidden core-n
states are projected out before diagonalization, matrix elements H and I take a more
complicated form, which can be found in [65].
In actual calculations, it may be useful to move between the Y and T Jacobi sets.
For example, the wavefunction in Eq. (2.21) is written in the T basis, but the matrix
elements of the core-n interaction are most simply calculated in the Y basis. A change
of Jacobi systems only affects the spherical and hyper-angular parts of hyper-harmonics
25
twill
in E9 (2'15.
3 unitar." ire!
,2 ’
3m! JET“,
when? the: pri
-’ (l‘ lllll’l'
in!‘ l” J '
the Rayneil-llt
f)fi’1}'[)t’l‘-lltll'll£
Clt‘fliS (“an ln'
' I I _. .
All. y.1...\. ll
59f Fur t'xt'iirm
by Eq (2.“- I
are essentially :
higher partial u
C’irt’ffit‘it‘rits will
To (‘Ullt‘lllili'
betwwn gm mu -r
the (1 ire in Eq. 1".
be dealing with :-
.l
ow nucleons ii
SW .
,ttm. an opera
L'J‘fll center of in
when ‘7
» F, Would be t
:0“.
‘ .er -
1 (root of tin
' I
in Eq. (2.18). Under a transformation of Jacobi sets, hyper-harmonic functions undergo
a unitary transformation:
W’yval Jgal (B’Qx’ay’xnl’XRZ): Z (I?!) l, yll$ly>KLg’ Jflal (0, chiQLaanXnQ) a
at
1;. 1.’
(2.27)
where the primed and unprimed quantities refer to different Jacobi sets. In 7m] and
720,, quantum numbers K, L and S are the same. The coefficients (1’ l’ lely) K L’ called
the Raynal-Revai coefficients [100], have analytic forms. Because of the orthonormality
of hyper-harmonic functions in all quantum numbers, squares of Raynal-Revai coeffi-
cients can be interpreted as probabilities to find a basis state with quantum numbers
K, lg, lg), L, S in one Jacobi set in a state with numbers K, 13;, lg, L, S in the other Jacobi
set. For example, for K = 0, there exists only a single combination la; 2 lg = 0 allowed
by Eq. (2.8). Therefore, in Eq. (2.18), basis states with K = 0 in different Jacobi sets
are essentially identical. Also, low partial waves 13,13, in one Jacobi set may contain all
higher partial waves 1' ,l’ allowed by Eq. (2.8) in the other Jacobi set. The Raynal-Revai
coefficients will be useful in Chapter 5 where 6He will be scrutinized.
To conclude the discussion of the three-body model, we establish basic relationships
between geometrical measures within a three-body system. We assume, for a while, that
the core in Eq. (2.21) is described by a microscopic wavefunction @Jgore' Then, one would
be dealing with a system of A nucleons consisting of Z protons and N neutrons. The first
Acme nucleons including all protons would be contained in the core. For such an A-body
system, an operator of the average squared distance of nucleons from the position of the
total center of mass TC M S — (1 /A) 2,4121 1*",- could be defined as:
A
n—5 AZ” rz'—7‘CMS)2 , (228)
where 1",- would be the position of the i—th nucleon in the laboratory frame. After taking the
square root of the mean value of this Operator applied to the wavefunction in Eq. (2.21),
26
*3.-§'
one would ()l
of the nllt‘lt‘l
where l'lé'llt‘
of the cure re
with the sun;
The C(Jfft‘.\[i(i
.‘ j -)
“H61? (FL ('(Il
distance lwt \\
FEM”) all (1]
With The
9
n Wmfld [
Sum]
i"
)0 r.
In {pamv
(3!" (I.
mfjdffl. P
"“llli'r.
I '4
‘h‘ thrree. m1
\.
one would obtain the commonly computed root- mean- -square (rms) matter radius (r )1/ 2
of the nucleus, which can be expressed as:
(rial/2 = fivmeagjcm» + (12%]. (2.29)
where () denotes an expectation value and (r?n(co're)) is a square of the rms matter radius
of the core relative to the core’s center of mass. Similarly, one could define an operator
with the summation over protons only:
1 z
223-3226401143)?- (2'30)
Zi=1
The corresponding rms proton radius (7%)”2 of the nucleus could be computed as:
(7,2,)1/2 = \flrg(core)) + (rgm_CMS>, (2.31)
where (r12,(core)) is a square of the rms proton radius of the core and rme_c M S is the
distance between the core’s center of mass and the center of mass of the whole nucleus.
Finally, an operator for the neutrons could be defined as:
N
1
T7215 N 21(7) '— TCA/IS)2 (2.32)
with the summation restricted to neutrons only. The expectation values of 7%,,1‘2 TP, and
7% would be related simply as:
1
<73.» = 2 [20%) + N630] (2.33)
In reality, however, the core in Eq. (2.21) is a macroscopic object. Therefore, the
core’s rms matter and proton radii can not be directly computed in the three-body
model. Rather, they must be inserted into Eq. (2.29) and Eq. (2.31) by hand. Within
the three-body picture, all protons in a two-neutron halo nucleus are confined inside the
27
Sf‘atte
core It is N
that tin? fl!)
2.3 “1
A> part of 1
in this that:
pFPSI‘ilUWl in
:10]: was 11w
in the pain-r
neutron svpu
was finally ('1.
change the (ll
energ}:
2.3.1 Int
hi the early (l.
to describe pi
the COW + n 1
effective 9I l 1
‘- l
models. two 111
Of the rah-11W
interaction 11M-
(‘OIM and n~n
and emphasis v
i . -
m. expanded
[lf'a ‘
amt llll‘t‘ff~l)()fl
ring.
Early llllf'r(),\
core. It is merely due to the core’s motion relative to the center of mass of the nucleus
that the rms proton radius of the nucleus in Eq. (2.31) is larger than that of the core.
2.3 11Li in the three-body model
As part of the present work, 11Li was studied within the three-body model introduced
in this chapter. The 9L1 core is allowed to be deformed and/or excited. The material
presented in this section is based on the article [50]. In the paper, the value 295 :l: 26 keV
[101] was used for the two-neutron separation energy in 11Li. However, as also commented
in the paper, a new experimental value 376 :1: 5 keV [102] was reported for the two-
neutron separation energy after the completion of our calculations, Later, this new value
was finally corrected to be 378 j: 5 keV [14]. In this section, no attempt has been made to
change the discussion and results to account for the change in the two-neutron separation
energy.
2.3. 1 Introduction
In the early days, three-body models of two neutrons and an inert 9Li core were developed
to describe properties of 11Li [13,58,61,103]. At that time nothing was known about
the core + n subsystem 10Li, and theorists could play the game of adjusting freely the
effective 9Li-n interaction in order to produce a sensible 11Li ground state. In these
models, two neutrons were coupled to the ground state of 9Li and the final composition
of the valence part of the wavefunction varied significantly depending on the core-n
interaction used. In [58] a three—body force was introduced in addition to the two-body
core-n and n-n interactions. In [61,104], a density-dependent n-n delta force was used
and emphasis was given to the importance of pairing. Three-body inert-core models have
been expanded to generate three-body continuum states [105] and the complexity of
these three-body scattering states was analyzed within the context of proton inelastic
scattering.
Early microscopic calculations were unable to reproduce a realistic binding energy
28
git
for 11Li wit
l’ldd bw‘ll l‘l
pairing ('lh”
within the r
nearllHJ1er
(finfigurathn
dueed h)" ll"1
eleven [‘37. T
mt‘xlels still
There an
state only. S
structure of
these i.\'>tit‘.\ 1
state. the 1‘
its low ext-it.
extenilml 111.,
state of m.- .
the ground s
first 0Xt'itwl
Slgtlifit-ant in
differing {pm
There, ([1,. gr
N‘sppm l\'t ]\
9-‘Ifitod
12
B
Slate
(3. Fur “71.
We 01in pr“!
for 11Li without artificially renormalizing the interactions [81,106]. As pairing effects
had been identified to be crucial [61], by introducing a phenomenological force in the
pairing channel, a self-consistent description of the Li isotopic chain became possible
within the relativistic Hartree—Bogolyubov framework [107]. Effective interactions valid
near the driplines have meanwhile been developed in the shell model [108]. Nevertheless,
configuration mixing, required to produce a realistic ground state for 11Li, is still intro-
duced by hand. In the mean time, some ab-initio methods have reached nuclei with mass
eleven [67,70]. Although the general spectra for light nuclei look promising, ab-initio
models still have difficulties dealing with halo nuclei, 11Li in particular.
There are still some open questions regarding 11Li, even when considering the ground
state only. Should the excitation and quadrupole deformation of 9Li play a role in the
structure of “Li? Using the three-body framework, it was our aim to shed some light on
these issues under the constraints provided by the new 10Li data. Besides the 3/2“ ground
state, the 1/2‘ first excited state of 9Li should also be present in the model space due to
its low excitation energy of 2.69 MeV [99]. We indeed carried out calculations in such an
extended model space, but results were not sensitive to the inclusion of the first excited
state of the core. This is probably due to the lower spin of this state when compared to
the ground state; there are no new orbitals brought into the configuration space when the
first excited state of the core is included. We then concluded that core excitation is not
significant in 11Li. This conclusion is in contrary to 112Be (2 10Be + n + n), a nucleus
differing from 11Li by a single proton, which has been studied within the same model [92].
There, the ground state and the first excited states of the 10Be core have spins 0+ and 2+,
respectively. Consequently, the model space of 12Be is enlarged by inclusion of the first
excited state of the core, and as expected, core excitation was found to be important in
12Be. For reasons given in this paragraph and to avoid unnecessarily tedious discussions,
we only present results for 11Li built on the ground state of 9Li.
29
9.1
ll)
2.3.2
. In the three
a ngl that ‘1
11All is 51211111
that have in.
energy. parit
Within tl
above the ”I,
by 121]. wher-
the (T‘t‘llte‘rq ,1".
'5 : 1 2 l)i"lli_
a UHF—n Put:
the mni + 11
Structure of 1'
1- Tllt‘ gm
450 k..\
2. there i5 ,
p3““‘9111.
3- Them is
[Thin-ml I
Along
' With Utli
melt-W ion.
2.3.2 10Li
In the three-body picture, one can not understand the Borromean nucleus 11Li without
a good description of its particle-unbound core + n subsystem 10Li. The information on
10Li is summarized in [99]. Therein, it is possible to see the large number of experiments
that have been performed to measure the spectrum of 10Li, but also the contradictory
energy, parity, and spin assignments made.
Within the few—body picture, 10Li is considered as a core + n system with a neutron
above the 9Li core. The extra neutron is allowed to live in orbits around the core labelled
by nlj, where n is the radial quantum number, I is the orbital momentum relative to
the center-of-mass of the core, and j = I <8) 3 is the neutron’s angular momentum with
s = 1 / 2 being the neutron’s internal spin. These orbits are assumed to be produced by
a core-n potential to be defined later. Based on the experimental evidence and using
the core + n decomposition, one can conclude the following regarding the low energy
structure of 10Li [99]:
1. The ground state of 10Li contains a valence neutron in a 231/2 state at about
+50 keV or below.
2. there is a 1191/2 resonant state at several hundred keV, to be also referred to as the
p-resonance. This resonance is often assumed to be around +500 keV [58].
3. There is no clear evidence for a d-state (l = 2) below +3 MeV. This state will be
referred to as the d-resonance.
Along with other constraints, these observations are used to fix parameters of the core-n
interaction.
2.3.3 Interactions
In this section, the potentials appearing in Eq. (2.22) used to bind 11Li are discussed.
In three-body models, the interaction between the two neutrons, V"‘"‘, is usually taken
from a parameterization of the low energy nucleon-nucleon scattering phase shifts, which
30
..a'
are well 11ml
the CPT im
The qttatimt
.\lU.‘l Hill
the 9Li m ’1'."
two pair-Wis.
form (if flw (
adefornmi h
by a spin—orb
potent ial:
t't‘or.‘ ~71 , a
i I'
(‘f’
where I is the
the operator «:1
{h (my)? :
The contra:
formation .i3 1
are spherical a
Rm : 1.2.3.41
“hen 32 # 0. 1
tion lmpost
d
«l on
(formation is x.
{012 2 0.7. a \‘n
Radius 5,, was 11
Standard value 1
the depth 1,}; of
are well understood. For V"“”, we use the soft-core Gogny interaction (also known as
the GPT interaction) [109]. It includes central, tensor, spin-orbit, and spin-spin terms.
The quadratic LL term is neglected since its effects are not strongly felt.
Most ambiguities reside in pinning down the effective interaction che"" between
the 9Li core and each neutron. In the following discussion, Fem-8-7,, stands for any of the
two pair-wise coordinates between the core and the neutrons depicted in Figure 2.2. The
form of the core-n interaction is based on the idea of the core being a rotor generating
a deformed field. This field is taken as a deformed Woods-Saxon potential accompanied
by a spin-orbit part proportional to a derivative of another non-deformed Woods-Saxon
potential:
—1 2
Vcore—n (Foam—nag) 2 thJS [1 + eXp (Tome—n _ R(0,¢))] _ (L) (2-34)
aws mnc
(21". 5') V30 d [1 + exp (Tcore—n _ R30)] _1 ,
47'core—n drcore—n 030
where f is the operator of the orbital momentum between the core and a neutron, .§' is
the Operator of a neutron’s spin, and m7, is the mass of a pion. For practical calculations,
(Ii/(m.,rc))2 = 2.0 fm2. Both angular momentum operators are in units of h.
The central Woods-Saxon part in Eq. (2.34) depends on the core’s quadrupole de-
formation fig through the radius R(0,¢) = M3(fig)[1+fing0(0,¢)], where 0 and (b
are spherical angles in the rest frame of the core. When fig = 0, a standard value
Rms = 1.2514253} = 2.60 fm is used [110], where Acme is the mass number of the core.
When fig 75 0, the radius parameter Rw3(fig) is adjusted to meet the volume conserva-
tion imposed on the central part of the interaction [10], and the dependence of Rms on
deformation is shown in Figure 2.3. In this and other figures, results are shown only up
to fig = 0.7, a value that is unrealistically large. The spin-orbit term is left undeformed.
Radius R30 was made equal to Rms at any deformation. The diffusenesses are fixed to the
standard value aws = ago = 0.65 fm [110]. To increase the flexibility of the interaction,
the depth V123 of the central part depends on the relative orbital momentum 1. Different
31
Baal
Figllf
lid; and lily;
all partial “m
The rut: lf‘
impose that tl
between the g1
known. Thvrvl
free paraimter
parattleter ('em
ferre for (JlJlatt'
and therefore l6.
( .
0f the 3L1 (‘Urf‘
a prolate (ll-furl
9m [111].
la (‘85? of an
orbital are (legr-
srnmetry is lm
longer goorl qlm
Oi (llfir
:rent '
n], r
Nevertheless. fur ‘
2.60 "' .No
2.58 - \
2.56
/
Uflfl
‘0 2.54
I I
A l I l
Rw
2.52
I
C
2.50
0.0 A 0.2 i 0.4 A 0.6 0.8
‘32
Figure 2.3: Radius of the 9Li-n interaction as a function of deformation.
V538 and V53 are considered for l = 0 and 1, respectively, and the same Vgs is taken for
all partial waves with l 2 2. The depth V30 of the spin-orbit term is l-independent.
The rotor-like picture for 9Li is probably an oversimplification, and thus we will not
impose that the quadrupole deformation be determined by the strength of E2 transition
between the ground state and the first excited state of the core, which in any case is not
known. Therefore, in the three-body model of 11Li the deformation parameter fig is a
free parameter chosen from a physically reasonable interval. In principle, the deformation
parameter can take negative values if the core is oblate. We found that the quadrupole
force for oblate shapes of 9Li produces more repulsion when compared to the prolate case
and therefore less binding energy. For this reason, only the prolate deformation (fig > 0)
of the 9Li core is considered in what follows. It should be noted that the preference of
a prolate deformation contradicts a recently measured negative quadrupole moment of
9Li [111].
In case of an undeformed 9Li, all J 7" states in 10Li originating from a given 711,- neutron
orbital are degenerate. The degeneracy is, however, removed as soon as the spherical
symmetry is broken by non-zero deformation. When there is deformation, l, j are no
longer good quantum numbers, and a nuclear state J'”(10Li) contains a superposition
of different 721,- components coupled to the ground state of 9Li with J“(9Li) = 3/2".
Nevertheless, for simplicity, we will refer to any multi-component state J"(10Li), by the
32
git
m’; cornptm“
The (lt‘l’l
in 1ULi. B” i‘
positions Hi ‘
{fibulliillit‘ I"
we plztt't‘ tltti
llIlL‘ltEllS with
be ('ltN‘ in +
(l—ftNtlié-illt't‘ \
not he re‘pl‘m
the p-l'f‘Nllldll
final restrit-tit
orbital at - 1
assumptions.
interartion prt:
of zero def. in”
deli”lb 50 that
+30 keV. 4. ll)
di‘lf-Jrniation i§
and the lowm
“9 no“: lllélf .
rmlwl to term
Finally. in r
1.. .
P32 “rum.“ ,
nlj component into which the state collapses in the limit of no deformation.
The depth of the core-n interaction can now be adjusted to reproduce low-lying levels
in 10Li. Because the experimental data on 10Li from Section 2.3.2 is inconclusive about
positions of the p— and d-resonances, these states can be moved around a bit to obtain
reasonable results for 11Li. In particular, to reproduce the experimental binding of 11Li,
we place the p-resonance in 10Li at +400 keV. By comparison of 10Li with 11Be, a
nucleus with the same number of neutrons, one would expect the d—resonance in 10Li to
be close to +2 MeV [112]. In this work, the d-resonance is placed at +3.4 MeV. If the
d-resonance were at much higher energy, the three-body binding energy of 11Li would
not be reproduced; if it were at much lower energy, it would become bound more than
the p-resonance at large deformations in contradiction to experimental data on 10Li. The
final restriction on the core-n interaction is that the interaction must produce the lp3/2
orbital at —4.1 MeV to match the neutron separation energy of 9Li. Under all these
assumptions, the depths VJ”, and V30 of the core-n interaction are adjusted so that the
interaction produces a series of levels, lowest of which are shown in Figure 2.4 for the case
of zero deformation. In the deformed cases, fitting the core-n potential means adjusting its
depths so that the centroids of 1123/2, 231/2, 1121/2 and 1d5/2 orbitals are kept at -4.1 MeV,
+50 keV, +400 keV and +3.4 MeV, respectively. The variation of potential depths with
deformation is finally shown in Figure 2.5. The corresponding two-body bound states
and the lowest resonances in 10Li are shown in Figure 2.6 and Figure 2.7, respectively.
We note that, with this choice of the core-n and n-n potentials, no three-body force is
needed to reproduce the three-body binding energy of 11Li.
Finally, in order to approximately satisfy the exclusion principle, the bound 131/2 and
1133/2 neutron orbitals are projected out of the model space before diagonalization [65].
2.3.4 Results
Calculations were performed using the computer code EFADD [113]. The full model space
in Eq. (2.21) contained all valence terms with mag S 18 and K S 22. The hyper-radial
33
..-;il
figure 2.4: T
asurnwl that
that the "Li .
were used to I
and 11):; .3. are
valente in: ulwl
Laguerre fnnrt
Figure 2.8 .~
size of the nn n it
three sets of FPS
Exhibits the m:
Studied. ()n the
than in ”,0 12,
Flgllre 2:}
I ,
E2body nlj JWOOLi)
+3.4 MeV —-‘—— 1d5/g 1:213:4-
+400 keV — 1191/2 1+,2‘L allowed levels
+50 keV — 231/2 1-,2- __
-4.1 M v — 1 0+ 1+ 2+ 3+—
e p3/2 ’ _’ ’ forbidden levels
— 131/2 1 ,2 _
Figure 2.4: The lowest energy levels in 10Li produced by the core—n interaction. It is
assumed that the valence neutron is coupled to the J"r = 3/2- ground state of 9Li and
that the 9Li core is undeformed, i.e. fig = 0. The levels, energies of which are shown,
were used to restrict parameters of the core-n interaction. The two lowest levels, 131/2
and 1p3/2, are forbidden for valence neutrons in 10Li and need to be projected out of the
valence model space. The other orbitals are free to be occupied by valence neutrons.
Laguerre functions reach out to the maximum hyper-radius 20 fm.
Figure 2.8 shows the convergence of the three-body binding energy of 11Li with the
size of the model space measured by the maximum hyper-momentum. The figure contains
three sets of results corresponding to fig = 0.0, 0.3 and 0.6. On one hand, the convergence
exhibits the well known exponential dependence when Kmax 2, 12 for all deformations
studied. On the other hand, in all cases, the convergence rate is very slow, much slower
than in the 12Be case [92]. Moreover, the convergence rate decreases with increasing
-25 4+ T +
3
_.—v
W8
-30F ,x""x
t -X’ - o--vp
-35- --—-X“x’ “'8
S. b x-._aX—-—-X ...‘-.-v:s
O 40F o"'O""O“’O"-O——-Q__ ....x...
2 . 'O"‘0 V80
> 45
.50; ‘....A-...‘..__‘“”".”‘... -
t = = = : = = - -
-55
Figure 2.5: Depths of the fitted 9Li-n interaction as a function of deformation.
34
git
fianre 2.6: T\
Thy ngf‘xul 1"
9th and the t
(leftiirrnat ion. l
energicis.
The three-
in Figure 2.9.
h
.WIII : 22 m
.. . 1"
Case of *Be wl
ohtained from t
Figure 2.7:
Wilton. Tilt? lf-tu.
state of 9 m
i) the
the other states I
t , ....t-"""""'1p ,0+
.al- ”9*. * 3,2
-4 _' Ma" '-§:2:’;‘"-"~6=---.6:':;§:'-’7—§ ‘ ‘V‘ ' 193/2' 1+
%. -5: ‘NE‘.. in 1p 2+
2. “6.; m *
g-za- Mme
N .
m .24 b / —I—131/2p1
’ ...,o/O - 0- 1s 2-
-25 9+1 4 - . - 1 4 m.
0.0 o.2 0.4 0.6 0.8
$2
Figure 2.6: Two-body binding energy of bound states in 10Li as a function of deformation.
The legend refers to nlj valence neutron orbitals (coupled to the 3/2“ ground state of
9Li) and the total spin of 10Li.
deformation. It is thus necessary to use extrapolated energy values for three-body binding
energies.
The three-body binding energy of 11Li as a function of deformation is presented
in Figure 2.9. The figure contains both the values for the maximum hyper-momentum
Kmax = 22 and those obtained through the extrapolation in Kmax- Contrary to the
case of 12Be where the energy gain was large, in 11Li a very small additional binding is
obtained from the quadrupole coupling, and as the deformation becomes large the system
6 ’ I f l I . *_.—1d5R'1
5: r ...-"'r ’4'1d5,2,2
"' *I"° .. .. -
4 i _. *'='.L.'_-;i;.-_:§;_—_§;-_ _ _ ...- _ ... t 105,2. 3
S. 3 :- .x---.x-'-.x-'x-'1dy204-
O 2 ..
a 1" H." 1P1”, 1+
3‘ 0.6 ‘ b... A +
. .. A-"' -. _.
ul“ 04 - a--.o...-_-.e-..1"""" v ”V22
. . v-._.v_.-.v‘.‘. —'—28 .1-
0 2 _ v~.-_v 1/2
. L Yao-fl-mrn-gr-QT _—"-8l- o- 281/22-
0.0 ‘ ‘ '
0.0 0.2 0.4 0.6 0.8
I32
Figure 2.7: Two-body binding energy of continuum states in 10Li as a function of defor-
mation. The legend refers to 711,- valence neutron orbitals (coupled to the 3/2‘ ground
state of 9Li) the and total spin of 10Li. Both s-wave states refer to virtual states while
the other states correspond to real resonances.
35
...-a'
{J—
Flg‘lt‘t‘ 2.8: De
spat-e deterrni
any Km“. all
The lines are t
heroines has it
The rtns tn.-
compute the in
mm. (2.20". Ill
rnimentntn lx’,,|
less than 1'7. ll
Impose a ('Utistt
that the extrapc
observable that
for the
rrns 1th it
the ment tIlO‘dSl
Figure 2.11 sl
ground stat e of l
St-iieme /' .
‘ITJI “’51..
310m . .
.Illd along .I
r ,
Jl neutrons sittir
flit? er 1 '
..th, is small.
Figure 2.8: Dependence of the three-body binding energy of 11Li on the size of the model
space determined by the maximum hyper-momentum K max included in calculations. For
any K m, all possible valence channels with K s K max are included in the wavefunction.
The lines are exponential fits to the tails of data sets.
becomes less bound.
The rms matter radii of the corresponding wavefunctions are shown in Figure 2.10. To
compute the matter radii of 11Li, the rms matter radius 2.32 fm of the 9Li core was used
in Eq. (2.29) [13]. In Figure 2.10, the matter radii were obtained for the maximum hyper-
momentum Kmaa; = 22, but their variations between Kmax :2 20 and Kmam = 22 were
less than 1%. The experimental three-body binding energy and the rms matter radius
impose a constraint on values of the deformation parameter, namely fig ,3 0.3. We note
that the extrapolation was only done for three-body binding energies since it is the only
observable that has a well established exponential dependence on Kmax. Our prediction
g>1/2
for the rms proton radius obtained for fig = 0.3 is (r
p = 2.37 fm in agreement with
the recent measurement of the charge radius of 11Li [48].
Figure 2.11 shows the probabilities to find the three main structural components in the
ground state of 11Li, namely (31/2)2, (131/le and (d5/2)2 components. Here, j j coupling
scheme (livjlew'y) in the Y Jacobi basis is used (see Figure 2.1): la; and Z3, are orbital
momenta along Jacobi vectors {if and 37, and j ’s are orbital momenta l coupled to spins
of neutrons sitting at the ends of corresponding Jacobi vectors. When the deformation of
the core is small, the ground state of 11Li is almost 60% (p1 /2)2; for large deformations
36
figure 2.0: 'l‘l
correspond t:
the extrap: :En
(Ufrt'fiptiltth It
it beetlitlt‘s nn
ttansttiun l‘wtx'
eqtal prulmlilll
is small. lexs tl
lOT 1330 in :03
State of ”at.
‘ llrn~
respond t all“
{T "
..itrtsponr s to ti
0.1 T T r I I
. .’./.
0.0 -
%‘ -0.1 , «
a - o
E-OZP / v0”‘0~--o” q
(0 l I I
LLI i : 1‘ .2
-0.3
We“ . . a . .
'0'4 0.0 0.2 0.4 0.6 0.8
32
Figure 2.9: Three-body binding energy of 11Li as a function of deformation. The squares
correspond to values obtained for Kmaa; = 22, the circles are values obtained through
the extrapolation in Kmax for Kmax 2 12. Lines are to guide the eye. Shaded region
corresponds to experimental value 295 :l: 26 keV [101].
it becomes more than 80% (31/2)2. The region around fig = 0.3 corresponds to the
transition between these two configurations where both components are populated with
equal probability. Regardless of the deformation, the weight of the ((15/53)2 configuration
is small, less than about 7%. This result is in contrast with the three-body calculations
for 12Be in [92] where the ((15/2)2 configuration accounts for about 30% in the ground
state of 12Be.
4.3
f.)
4.1 /
.— 3.9
E
Q 3.7
A
“LE 3.5
v
3.3
3'1 0.0 0.2 0.4 0.6 0.8
32
Figure 2.10: Rms matter radius of 11Li as a function of deformation. The squares cor-
respond to values obtained for Kmax = 22, the line is to guide the eye. Shaded region
corresponds to the value consistent with reaction data [54].
37
..al
fitting the (c
Figure '3-11: I
of NU "L‘ a {i
One has t:
n interaction.
the strut-titre i
Rather. the m
1DLi. Other in
parainvtvrfi. P1
It is itnpnt‘
the large mini
stun wit it \Vi tit.
assumptions it
at +30 hi‘V: l
is the 2s1 .2 M
Changed by m
certainty in th
d‘TWHl‘dnf‘é‘ at
“$08M by r )1
d‘.‘TS()llaIl('(_‘ is '
Hi I‘
1.0 . ' . ' r ' T v
0.8 ~ """""' .
b O-- - -o- - —I— 2
g 0.6 . ‘C\ \ (81/2)2
'8 0.4 h \o — O- (p1/2)2
a I |——’ \\ * (d5/2)
0.2 - \ -
i*....*....*....*_ \\
L i L . ... 7""‘g”’1“'-8
0'0 o o o 2 0.4 o 6 o 8
‘32
Figure 2.11: Probabilities of the three main structural components in the ground state
of 11Li as a function of deformation. The lines are to guide the eye.
One has to realize that there is no unique parameterization for the effective core-
n interaction. We have convinced ourselves, though, that the features shown here for
the structure of the ground state of 11Li do not result from a specific parameterization.
Rather, the main features emerge from the constraints imposed on continuum states in
10Li. Other interaction parameterizations, using different interaction radii or spin-orbit
parameters, produce exactly the same characteristics of 11Li.
It is important to understand the implications of the 10Li structure on 11Li. Despite
the large number of experiments, a close study of [99] raises questions about the preci-
sion with which states in 10lLi are known. We have explored the possibility of different
assumptions for the neutron states to which the core-n interaction is fitted, namely 231/2
at +50 keV; 1121/2 at +400 keV and 1d5/2 at +3.4 MeV. Of these, the least uncertain
is the 251/2 state. We have checked that the main features of the present work are not
changed by moving the 1191/2 neutron orbital to +500 keV. More important is the un-
certainty in the location of the ld5/2 state. There is no clear experimental evidence for a
d-resonance at +3.4 MeV or at any lower energy. If the d-resonance is broad or it is su-
perposed by other states, it could be hard to observe experimentally. What happens if the
d—resonance is pushed down? We refitted the core-n interaction for deformation fig = 0.3,
fixing the centroid of the d-resonance at +2.5 MeV. An immediate consequence is a gain
38
e~al
of .\
9 ~.
tinfflbudv papf‘l'
fun-r
in threcb‘ “13
mt be suffi‘ll'
exw‘nIiRH-Y if:
23.5 CO]
We perform: *t i
and excitatiur.
amount for I}.
a threw-body
:0 rcprmlurc '
radius. and [is
excitation is fr
in 11Li is pmla
(‘dii'lli‘dllLIIlS 3t
theposiiion of '
that such a rm
State at a much
dstatcs in “’Li .
an exporinwnt 5
II. this Sfl'llUll. v.-
a(.{, .n», .
. {as to thrift-J“.
me outcome
Moms in filln ‘
in three-body binding in 11Li of about 150 keV. However this additional attraction would
not be sufficient to change the structure of the ground state of 11Li, which would remain
essentially (31/2)2 and (p1 /2)2, with ((15/2)2 accountng only for about 10%.
2.3.5 Conclusions
We performed three-body calculations for the ground state of 11Li including deformation
and excitation of 9Li. We find that reorientation effects due to core deformation can
account for the known configuration admixture of s—waves and p-waves in 11Li. With
a three-body model, in which the core is treated as a deformed rotor, it is possible
to reproduce the three-body binding energy, the rms matter radius, the rms proton
radius, and the structure of 11Li consistent with experiment. On the other side, core
excitation is found to be unimportant. In the three-body model, the strength of d—waves
in 11Li is predicted to be very small (x 7%), which is in disagreement with shell-model
calculations [50]. So far, experiments have not been able to make a clear statement about
the position of the d-resonance in 10Li. In the three-body calculations, we have assumed
that such a resonance would be above +3 MeV; however, the shell model produces this
state at a much lower energy around 2 MeV [50]. Resolving experimentally the position of
d-states in 10Li will settle once and for all the structure of the ground state of 11Li. Such
an experiment should be done with a reaction starting from 9Li rather than knock-out
from 11Li since there is not much d—waves in 11Li. One possibility would be to repeat the
9Li(d,p)10Li experiment [114] at a higher beam energy.
2.4 6He in the three-body model
In this section, we repeat some calculations for 6He published in [94]. The goal is to gain
access to three-body results for 6He and have them available for a comparative study with
the outcome of MiCH for this nucleus in Chapter 5. The main purpose of the original
three-body paper [94] was to investigate three-body continuum structure and response
functions in 6He, but as part of the work the wavefunction for the ground state of 6He was
39
Baal
:1
3i
‘1 -
construvtml 2'
lines dostrrlu
is regarrgiwl :l.
2.41 111'
The nuvle'us ‘
the years a: if
:atkiwi in a \“
cluster :76. TN.
rure of “HP. 1
ground stem".
€1‘P1’.’d§~ will in
between (iiife ‘11
In the tim-
lh‘dn 11Li. In t}
Hlfllfiti‘i Would gr
Well bound am
and the first (ix:
POmt 0f \‘ir‘w, '
mirlmns in tlw
from Section ‘2.
to assume that
contradict (“Ullf‘
1
Has been found
4 ~I
are He core on
bGCOI' '
Miiicmh'
srrr; '
h thing as a ii
suffers from um
I.
constructed as well. In that work, hyper-spherical expansion and Pauli blocking along the
lines described in this chapter were employed. The ground state of 6He in the cited work
is regarded as possibly the best within the formalism presented in this chapter [115].
2.4. 1 Introduction
The nucleus 6He, the lightest of the two-neutron halo nuclei, has been used throughout
the years as a reference nucleus in the realm of Borromean nuclei. Thisnucleus has been
tackled in a variety of models, ranging from three-body [13,94,116] through microscopic
cluster [76,78,79] to fully microscopic models [69,117]. Because of the fairly simple struc-
ture of 6He, many theoretical models are in agreement on the bulk properties of the
ground state, such as binding energies, radii and the occurrence of halo structure. How-
ever, as will be demonstrated in Chapter 5, one needs to look deeper to find discrepancies
between different models. Here, we focus on the three-body description of this nucleus.
In the three—body picture, 6He in its ground state is considerably simpler to tackle
than llLi. In fact, 6He is probably the two-neutron halo nucleus on which any three-body
model would stand the most firmly for several reasons. First, a free 4He is exceptionally
well bound among light nuclei, its J W = 0+ ground state has zero quadrupole moment,
and the first excited state (also 0+) is above 20 MeV [118]. Moreover, from the microscopic
point of view, 4He is a fairly simple object: in a first approximation, 4He contains four
nucleons in the lowest possible spherical 131/2 mean-field orbit. Here, the nlj notation
from Section 2.3.2 is used. Within the three-body approximation, it is then reasonable
to assume that 4He remains hardly polarized in 6He, even though this assumption may
contradict conclusions of some microscopic models in which distortion of the core in 6He
has been found important [76]. Thus, in three-body calculations of 6He, we consider
the 4He core only in its non-deformed ground state. Second, the core-n potential can
be conveniently fitted to experimentally known scattering phase shifts. But there is no
such thing as a free lunch: even with the core-n interaction fitted to scattering data, 6He
suffers from underbinding, as will be shown in Section 2.4.2. Third, ground states of both
40
klv‘ ____
ill:
4H9 and Gllr' .
Channels ()pr'n
the l«;»m’*>‘t J"
«:rnlv one that r
24.2 IIltC
Tilt“ IXR‘ri-lnirly '_
FUN -- 72
tion l' (”in ’
undefirrrrml. PM
R9,, = 1.3 fm. (1:,
[2,0 = —-«lll.(l .\lt“
atrium rerprr )(llll 'e -:
valence llt’lll rt iris
With these I“
m of undcr‘liirul
of insufficient tlir
m ~‘l
Pmiule ream ms
f‘lt’rsr‘d Clramwls.
Priiilr'rmials nr‘glm 'r
rrivrriZatiorr (rife-H
in? third ('lrrstvr I
ruf- i-
rrflrie tlirmuluu
:0 5i ' "
mimic the effl
4He and 6He are J7r 2 0+ objects which considerably decreases the number of valence
channels open for the two extra neutrons. Finally, the 1.91/2 neutron orbital (consequently
the lowest J 7' = 1/2+ state of 4He+n subsystem) produced by the core-n potential is the
only one that needs to be projected out of the valence model space.
2.4.2 Interactions
The two-body interactions from Eq. (2.22) are adjusted as follows. The 4He-n interac-
tion Vcore—n combines a central Woods-Saxon and a spin—orbit Woods-Saxon—derivative
parts as in Eq. (2.34). The nuclear field is spherically symmetric because the core is
undeformed. Parameters of the core-n interaction are taken from [94]: Rm 2 2.0 fm,
R30 = 1.5 fm, aw. = 0.7 fm, aw = 0.35 fm, V53, = 145,. = —43.0 MeV, V3,, = —21.5 MeV,
V30 2: —40.0 MeV. The core-n interaction is zero for partial waves with l > 2. This inter-
action reproduces a-n scattering phase shifts satisfactorily. As for the interaction between
valence neutrons V""", a realistic Gogny (GPT) force is used [109].
With these two-body interactions, the three-body model of 6He suffers from the prob-
lem of underbinding [13,116]. It is commonly argued that physics underlying the problem
of insufficient three-body binding may have to be explored beyond three-body models.
Possible reasons for underbinding include polarization of the core and the influence of
closed channels, most important of which is 3H+3H; use of local energy-independent
potentials neglecting exchange interactions that would be introduced through antisym-
metrization effects; and two—body interactions may not be the same in the presence of
the third cluster as those in a free space. To cure the underbinding problem in [94], an
effective three-body force was introduced:
v3b0dy= ‘1‘50 [MeV] (2.35)
1+(p/5.0)3
to simulate the effects of the closed 3H+3H channel.
41
twill
Fiatrro 2.12: ll
nlrir'lt’l 5pm 1“ (1t
tints. For ('éll'fl
wawhun t it in.
2.4.3 Rom
II. this With in.
rest of the Ilia '11
(“He are (“Hm
mmirlor a mm l
in the T .IdHJlll
Lit'lrrticd.
The corrvvrg.
Sartre drip
It
rmirm
'mqr : 4” ill!
:trbtainrxl from a
T ~ A
rhrralore. result:
For the (mm-1
flat-tibi basis: arr
”5‘5. d0 not lun
(ere srrtrrnml 0W '
amun: for r‘icar‘lx'
-o.6- ' .
Jl .
5. -o
o r
gig-0.8- ' -
_ _ ' exp=-O.97Mev
I]? 0.9 . \-l
, - .
-1o~ ' F' ' .
5 4 15 ' 25 l 35 1 45
Kmax
Figure 2.12: Dependence of the three-body binding energy of 6He on the size of the
model space determined by the maximum hyper-momentum Km included in calcula—
tions. For each point, all possible valence channels with K 3 Km” are included in the
wavefunction.
2.4.3 Results
In this section, some results for 6He studied in the three-body model are presented; the
rest of the discussion will be delayed until Chapter 5. Because the three—body calculations
of 6He are computationally cheap compared to the case of 11Li in Section 2.3, we can
consider a much larger model space than for 11Li. The wavefunction for 6He is written
in the T Jacobi basis, and all hyper-radial Laguerre basis functions with may 3 25 are
included.
The convergence of the three-body binding energy of 6He with the size of the model
space determined by the maximum hyper-momentum Kmax is shown in Figure 2.12. For
Kmaflc = 40, the three-body binding energy is -0.98 MeV, a value that would also be
obtained from an exponential fit to the high-Km“; tail of data points in Figure 2.12.
Therefore, results for Xmas: = 40 are considered converged.
For the converged state, the weights of the five dominant valence configurations in the
T Jacobi basis are shown in Table 2.1. As we explained in Section 2.1, hyper-radial basis
states do not have a well defined physical meaning. Therefore, the weights in the table
were summed over nlag for a given combination {K , 13;, lg, L, S } The five configurations
account for nearly the entire wavefunction. However, this group of components is by itself
42
.-al
git
Table 2.1: Pro
graurid state t-t
flares f and y“ t
taIPrIt-e part of
rat large (‘lll mull
when." the llll’f *1 ‘— l
mine. The? grtitrrt
singlet .19 = (II v.-
admixrd spirt-tri;
Sirttilarly. lllr“
ill Table 22 In t
r-= -
.rf'l
.15 with (I : l
Marts in agrm'n
5...! . .‘
..rt.lld\dli&1l)lf’ for
Figure 2.13 slit
.5.
.37.
'5
d. "
_ r1. fora givr'rr r
arrrlmtrltiplir‘rl l)\' 4
This form of hilt"!
Emit-5:: from Eq. ('2.
smii hrpcn rm 1 i i. l
1.52 ,. r'
trgtatrng sqlmrf‘rl
Flirting
T disr-uxxjr
Table 2.1: Probabilities of the five dominant components in the T Jacobi basis in the
ground state of 6He. Here, la; and ly are orbital momenta associated with Jacobi coordi-
nates 53' and 37 from Figure 2.1, L and S are the total orbital momentum and spin of the
valence part of the wavefunction, as in Eq. (2.18).
T Jacobi basis
alias K la, ly L S probability [%]
K = 2 s-waves 2 0 0 0 0 80.89
K = 2 p—waves 2 1 1 1 1 11.03
K = 0 s-waves 0 0 0 0 0 4.17
K = 6 d-waves 6 2 2 0 0 1.64
K = 6 f-waves 6 3 3 1 1 0.78
E = 98.51
not large enough to deliver converged results, as can for example be seen from Figure 2.12
where the three-body binding energy for K max = 8 is about 400 keV above its converged
value. The ground state of 6He is controlled by K = 2 valence components. The spin—
singlet (S = 0) valence configurations account for about 86.7% of the wavefunction; the
admixed spin-triplet (S = 1) components exist only due to a spin-orbit coupling.
Similarly, the probabilities of dominant configurations in the Y Jacobi basis are shown
in Table 2.2. In the Y Jacobi basis, the ground state of 6He is dominated by valence
terms with la; 2 1, i.e. by terms in which the valence neutrons orbit the core in relative
p—waves in agreement with the shell-model picture where the lowest p-shell is the first
shell available for valence neutrons outside the 4He core.
Figure 2.13 shows the hyper-radial dependence of valence configurations from Table 2.1.
Again, for a given valence configuration, hyper-radial dependences were summed over mag
and multiplied by a factor p"5/2
to bring them to the form u(p) introduced in Eq. (2.11).
This form of hyper-radial functions is preferred because of its simpler asymptotic prop-
erties from Eq. (2.14) and Eq. (2.15). As expected, all hyper-radial functions vanish at
small hyper-radii. Probabilities of valence configurations in Table 2.1 were obtained by
integrating squared hyper-radial functions from Figure 2.13 over the hyper-radius.
Further discussion on 6He is postponed until Chapter 5 where the three-body results
43
._;t|
(It
Table 2.2: Pr:
ground state t
l, are orbital
in Figure 2.1
VE’f‘l (KS.
figure 2.13: UV]
[[9 thrf‘fi‘])()(1v v
will
it. be (0111er «i
Table 2.2: Probabilities of the three dominant components in the Y Jacobi basis in the
ground state of 6He. Here, j j coupling is used. Schematically jz- = I, <8) 3i, i = x, y, where
Q are orbital momenta associated with Jacobi vectors (If and 37 in the Y Jacobi basis
in Figure 2.1 and s,- are spins of neutrons sitting at the ends of corresponding Jacobi
vectors.
Y Jacobi basis
15,; jg; ly jy probability [%]
1 3/2 1 3/2 85.01
0 1/2 0 1/2 7.83
1 1/2 1 1/2 5.79
2 = 98.63
o'st ‘ I . I i I ' j ' —K=2 s-waves
~ - - - - K=2 p-wave
0.4 » ------ K=0 s-waves
----- K=6 d-waves
H 0.3 - ------ =6 f-waves
Q .
'7 0.2
E .
:9: 0.1 _ . ................. .
3 0'0 a; -_,-’:,‘_‘_::..-.:..:.
-0.1- _
.. .2 - I ,‘T" l . l . ggl . I .
0 0 2 4 6 8 10 12 14 16
p [fm]
Figure 2.13: Hyper-radial dependence of the five dominant channels from Table 2.1 in
the three-body wavefunction of 6He.
will be compared with those obtained in MiCH.
44
.-al
at
.
)3 I
Chapt(
Core
I“ the pry-vii )ll.‘ H
was trtrtlirml. ll:
piti'pt‘t Br :rrt )Itlr' -.-:'
Wish to lllltl a stir"
tt'fptN'ttttttlUIt in
let us start ti
jZITJtpt‘ises of the pr
5‘33- ht‘i‘k'ever. is tr
have MiCH rlm't‘l‘ ’l
inscribe heavir *r ('u
iii‘: ' '
st. of running lItlU
that the size of (‘ill(
tatt. $0!th Hl()(lf‘l.\ .\
fitrrte' -'
..ts mth some i
[vi-’[Jiijflfd 1' ~
. rt c1.\-()lldl)l\' a
he '*
- \alerirre part ir-
Chapter 3
Core
In the previous chapter, the formalism for the valence part attached to a macroscopic core
was outlined. The three-body-like construction of the valence part should guarantee the
proper Borromean asymptotics of the wavefunction. To develop MiCH further, we now
wish to find a suitable microscopic model for the core to replace the core’s macroscopic
representation in Eq. (2.21).
Let us start the search for the core’s model by defining our needs. First, for the
purposes of the present work, we need a microscopic model of 4He. Our ultimate future
goal, however, is to tackle heavier two-neutron halo nuclei, especially 11Li. Therefore, to
have MiCH developed in a unified fashion, the model for the core should be capable to
describe heavier cores, such as 9Li. However, in the light of upcoming chapters, there is a
risk of running into computational difficulties for systems heavier than 6He. It is known
that the size of calculations in microscopic models grows rapidly with mass number. In
fact, some models such as Green’s function Monte Carlo are currently limited to masses
A g 12 [67] due to computational demand. For future applications, the model for the
core needs to be flexible enough to provide either a fully microscopic or microscopic
cluster picture of the core. In the later version, the core would be built from microsc0pic
clusters with some internal degrees of freedom frozen. Second, the core’s model should
provide reasonably accurate structure information so that we can focus on effects due
to the valence particles. This constraint requires a certain level of sophistication in the
45
-al
Qt
dd‘flpfilm ()f
nucleon-nut-lm )1
halo nut-lei lie
1.310071 to play .'
impose SpH'izrl 4
Cores are tight lj
r-t_»zr1pletel_\' in .l
{refittl'ttatex {till
rims motitm.
At first sight
4‘19l. 501110 Uf w];
Wt?!’t‘,"d])pli0d tr) 9
suits were f4: 01ml
1r;<;hrde: the Far l1
1
08:15 variational
mdel. the Crwrr
ration f1}'])("l'-.\‘])llt
lirChapter l and
-r.
0. the core, but tl
ti‘e '
._ requirertrerrt
stricture Illt le ‘ls‘
"l’ta't '~
9 "r L‘ (lt‘tlit'at
7 v
description of the core. Third, the model must handle central as well as non-central
nucleon-nucleon forces. This requirement is important because the studied two-neutron
halo nuclei lie outside of the lowest s-shell in the region where non-central forces are
known to play an important role. Fourth, unlike for the valence part, there is no need to
impose special asymptotic requirements on the core’s wavefunction, because all potential
cores are tightly bound systems. Last, the wavefunction of the core should be expressed
completely in Jacobi coordinates. Along with the valence part given in terms of Jacobi
coordinates, fulfilment of this constraint removes by construction the spurious center-of-
mass motion.
At first sight, the literature is very rich in microscopic structure models of light nu-
clei, some of which were mentioned in Chapter 1. In [119], several sophisticated models
were applied to 4He bound by the realistic AV8’ nucleon—nucleon force [120] and their re-
sults were found to be essentially the same. Models tested in [119] (see references therein)
include: the Faddeev-Yakubovsky method, the coupled-rearrangement-channel Gaussian-
basis variational model, the stochastic variational model, the hyper-spherical variational
model, the Green’s function Monte Carlo, the no—core shell model, and the effective inter-
action hyper-spherical harmonic method. Many microscopic structure models mentioned
in Chapter 1 and [119] meet some of the criteria imposed above on microscopic treatment
of the core, but the pool of models meeting all of them is very limited. The most restric-
tive requirement turns out to be usage of Jacobi coordinates. From all the mentioned
structure models, the stochastic variational model meets all our criteria. The rest of this
chapter is dedicated to this model.
3.1 Stochastic variational model
As its name suggests, the stochastic variational model (SVM) represents a variational
approach to many-body problems. SVM relies on the expansion of wavefunctions in terms
of functionally simple basis states and a stochastic Optimization of variational parameters.
The original idea of stochastic Optimization appeared in [121]. Later, the method was
46
..al
.P,.
imprm't‘d flll'l ll
djzierent. tht it 1.:
Sl'.\l has hwn
There are two '
the wat't‘fltttt‘t in
CIiSYM i> llN‘tl
In this St‘i'lit
I419. l5 Clt’N‘rllwt
rimglear systerris
Eur-iron interat'tj
Eiltfif‘ttll llllf‘l‘tu'll
39571 with tilt is
ffllt'ft"’r;-('l‘)t)l(“ ('ilhl
sxtems. and wit:
tr - 5
JACH. The lllH'l‘t
The accuracy
unit-tions. S\'.\l ]
Bettire their arlt i]:
few-body calt'ulat
Asa. the basis of
turn-ts and is (“m
In principle. .—
..wmerittnn I and
if." . .
...li)lfl‘dt if)“ “f ( '
I.
5here y
in} and }]
tar and a,“
improved further [74,122]. The basis functions used in the original and later version are
different, though. Hereafter, we shall by SVM refer to the later version of the model.
SVM has been also applied to a variety of problems in other branches of physics [123].
There are two versions of the model, differing by the treatment of the angular part of
the wavefunction: vector-coupled and global. Following [124], the vector-coupled version
of SVM is used in the present work.
In this section, a microscopic version of SVM applied to the core nucleus in 6He,
4He, is described first. This version of the model has been used to calculate properties of
nuclear systems with mass numbers less than eight bound by effective central nucleon-
nucleon interactions [74]. Later, the method was extended to include non-central nucleon-
nucleon interactions. PrOperties computed for 3H and 4He bound by non-central forces
agreed with those from other microscopic structure models [119,125]. There also exists a
microscopic cluster version of SVM, which has been applied to somewhat heavier nuclear
systems, and which potentially could be used to describe 9Li core in 11Li studied within
MiCH. The microscopic cluster version of the model is outlined at the end of this section.
The accuracy of any variational method crucially depends on the choice of trial wave-
functions. SVM prefers correlated Gaussians [126,127] as trial variational basis functions.
Before their adoption in SVM, these functions proved to be remarkably flexible in various
few-body calculations, mainly in atomic and molecular physics, see for example [128,129].
Also, the basis of correlated Gaussians allows analytical computation of many matrix el-
ements and is easily adaptable to the permutational symmetry of fermionic systems.
In principle, any one-body square-integrable function of vector 7'" and with angular
momentum l and its projection m can be approximated to any accuracy by a linear
combination of Gaussians of continuous size parameter a:
exp (—;a2) Mme), mm = r’thtaa, (3.1)
where 321m and Ylm are solid and spherical harmonics, respectively, and 0,. comprises
the polar and azimuthal angles associated with F. The radial factor in solid harmonics
47
.5”
Qt
improves the ~
rle particle ea:
or a correlated
wild consist H
apprt ,tar'h. int f -r-
SVM follow|
icilftji'fl (‘UUTI [111.11
Being rf‘lathe t-t
but also make If
hart-antisi'rnniet;
HUNT-1‘7.
Rh? 1 7
(ff JCUN (It‘llfl
l
”[Tr‘tflrf an) the ll
Fr
] l
\
afit)‘. little-x ~,
“l,
identifier K or if
P
H
]l]2]JI‘l2[‘{j-.] i\ (
f3Tf>t'tlirra[(.‘.‘.
improves the short-range behavior of basis states. This simple observation for a sin-
gle particle can be easily generalized to an A-body problem in either an uncorrelated
or a correlated fashion [130]. In an uncorrelated approach, the many-body basis states
would consist of Slater determinants of single-particle Gaussian packets; in a correlated
approach, inter-particle correlations are explicitly built into basis states.
SVM follows the correlated approach and employs basis states expressed fully in
Jacobi coordinates. Jacobi coordinates for a four-particle system are defined in Figure 3.1.
Being relative coordinates, Jacobi coordinates not only engage inter-particle correlations,
but also make the removal of the spurious motion of the center of mass trivial. Then, a
non-antisymmetrized basis term for 4He in either K- or H-like Jacobi basis is written as:
_. 1
¢7COT€ Jgore TCOTB NITCOTC (II), A) = exp (" 51:14.73) X
[611l2l3L12L (if) ® X5125123S]
11'
Jcore
TT12T123TcoreA/[TCOTC 1 (3'2)
where J3me denotes the total angular momentum and the parity of the state, Tom-e and
MTcore are the total isospin and its projection. The projection of Jame is suppressed for
clarity. Index 760"; comprises all other quantum numbers as well as the Jacobi channel
identifier K or H, i.e. "more = {11,l2,13, L12, L, 812, 3123, S, T12,T123, K/H}. The function
6,1,213L12L(f) is chosen as a vector-coupled product of solid harmonics of relative Jacobi
coordinates:
(Mamie) = [[37, (a) 5 372052)] L12 8 33053)] L. (3.3)
The spin x51231235 and isospin TT12T123Tc07‘e MTcoq-e parts consist of successively coupled
48
.5”
{it
4tvi..rlepartit‘lt‘ '
77.127‘123
The pfl'l_]t‘('ll( in r
tiers. Const rairu
Split-S [51-3. 5133.
to as different or
The Carissittr
3! 3tllitletl>lt tutti
earh basis elerm'r
Mgr
“nil .‘lrnrs : 4 ]
requzrerttetit. the 1
he llUllt‘t‘ that .
".'\ t-
Ly .0 Afar; _\ l .:
he. renter of mass
it sli
at]: r .'
d 51mph. n’l'ril
3:179 L7 IS
‘1 a trilllxi
single-particle spins and isospins:
X8123123S = [le <83 Xp2]512 ® Xp3]Sl23 <83 Xp4] S , (3.4)
69 71,4] . (3.5)
TCOT'BMTCOTB
7T12T123TcoreMTcore : [[[Tpl ‘3 71,2]le ® 3’3]T123
The projection number MTcore is fixed by the number of protons and neutrons in the nu-
cleus. Constrained by vector coupling, different sets of orbital momenta {l1,l2,13, L12, L},
spins {$12, 3123, S}, and isospins {T12, T123, Tome} may exist and they shall be referred
to as different orbital, spin, and isospin channels, respectively.
The Gaussian part in Eq. (3.2) deserves‘closer attention. The symbol A stands for a
3 x 3 dimensional positive-definite, symmetric matrix of non-linear parameters, specific to
each basis element. The quadratic form mAa: involves scalar products of Jacobi vectors:
Aawe—l
:rAcc a Z A,,-.5,- . 5,- (3.6)
i,j=1
with Accra = 4 being the mass number of the core nucleus. Due to the symmetricity
requirement, the number of independent elements in matrix A is Accre(Acore — 1) / 2 = 6.
Also notice that, due to dimensions of matrix A, the summation in Eq. (3.6) goes only
up to Acme — 1 = 3 which prevents the Awre-th Jacobi coordinate 554, the position of
the center of mass of the nucleus, to enter calculations. The Gaussian argument can also
be written in a slightly different form:
ACO’I‘B 2
32.41! = Z aij (Fpi — ij) (3.7)
i l
Titjh’S' lttshlt‘ lll
tit-peratt‘tr anti 1)
runzntatiun llltl
piained belt-1w.
present in the
Uni) it}? A. The
index.
The nuclear
T, and lltlt'lt‘titl-
.\-'1tethat the hi
LIFT" -
arise Stlt'll (m
9):“ -'
panston (‘Ut'lflt'
nation in Eq (0
ES (I) 2 Z
1' (4(1),-
Witl
l )
lt ‘ --
LS lnSll‘l](_-j]\.
(3.9). ’1
53-. In such a case, quantum numbers 11,12,13 in 9111213L12L123L (f) loose their meaning
of orbital momentum quantum numbers for a given basis state, and can be considered
discrete variational parameters, instead.
The full wavefunction is then written as an antisymmetrized linear combination of
basis sets from Eq. (3.2):
= Acore CA
700"? Jgo-re Tcore MTcore (157007? Jdrore Tcore AITcor‘e ' (
3.9)
Jgore T60” MTcore
Here, Acme is the antisymmetrization operator which runs over all permutations of par-
ticles inside the nucleus, i.e. Acme = Z? core!(—-1)”P with P being the permutation
operator and p being the permutation parity. The sum in Eq. (3.9) is left without any
summation index because the wavefunction is constructed stochastically as will be ex—
plained below. Due to the stochastic optimization, there may be several basis terms
present in the wavefunction with exactly the same combination of cycore and differing
only by A. Therefore, the linear coefficient c in Eq. (3.9) must also carry the A-matrix
index.
The nuclear Hamiltonian for the core nucleus includes kinetic energies of all nucleons
T,- and nucleon-nucleon potentials Vij:
ACOTB ACOT‘B
H: 2 71+ 2 Vij. (3.10)
i=1 1=i,', and energy and overlap matrix elements are computed as in Eq. (2.26)
with the Hamiltonian from Eq. (3.10).
It is instructive to visualize the effects of the antisymmetrizer Acme on the wavefunc-
tion in Eq. (3.9). To do that, a new set of vectors, a set of spots 7"},2' = 1, 2, 3, 4, is defined.
51
txal
TOT iht‘ Th:
:05. that
pfiflilfll’s El
(“ijfisi'tiflt‘ll-
3150 in Ft)
r pern
7». hut
Satin?” 2
Enough tl
ilviif tr.) {-'.
pie-rely exe
have \rtriisi
'part it it’s"
eiithe 111117:
In S\'.\l
tar rartrlc ur;
For the non-antisymmetrized system of particles, spots coincide with single-particle vec-
tors, that is 1", = FPi' When the antisymmetrizer is invoked, spots are not affected, but
particles are: in particle permutations, particles are moved around between spots. As a
consequence, each particle index in the definition of Jacobi coordinates in Figure 3.1 and
also in Eq. (3.2) needs to be replaced by an index of the spot hosting that particle in a
given permutation. This way, not only single-particle spins and isospins travel between
spots, but also Jacobi coordinates are changed. As an example, the permutation swapping
particles 2 and 3 causes the following reorientation of Jacobi vectors:
4 q
_, $1 1‘1
P1 332 P3 P1 P2 P1 P3 P1 P2
-' (-—) " -' f 3 _, H ...
3:1 53 P2 P3 3:2 5;}, 11:1 332 P2 P3 1123
P2 194 P3 134 P2 194 P3 52 p4
Through the distinction of particles and spots, the wavefunction automatically adjusts
itself to the action of the antisymmetrizer. Once the antisymmetrizer has been com—
pletely executed, spot indexes are the only ones prevailing, while all particle indexes
have vanished. Needless to say, particles are dummy objects. Yet in literature, the term
“particles” is used to refer to both our particles and what we call spots. For the clarity
of the upcoming discussion, however, we find the concept of spots useful.
In SVM, all variational parameters are optimized stochastically. The main arguments
for random optimization go as follows:
a The simplest choice of non-linear parameters would be to use only a diagonal matrix
A, elements of which could be chosen deterministically, for example in a geometric
progression [131]. The problem with such an approach is that the parameter grid
for each diagonal element of A would need to be dense enough which would re-
sult in wavefunctions with many terms. Even then, using deterministic methods of
parameter selection, it would be hard to avoid local minima in multi-dimensional
parameter spaces. Moreover, not all of the parameter grid points would be equally
important. The reason is that for a given set of Jacobi coordinates and orbital,
52
aal
fifi.".-‘~.-
0 As a rmm
spin. and
dé‘ll-‘p' “(if
repr(_‘_\('lll .
Thi> l)“ '1“
uf “(HI-ll“!
o Furtlwrmu
llt' (hN‘ripI
10 be“ (1)1153
1.’)fIlll('l(‘Hli\l
than n! lu‘rs
o Aim. in [W11
would. 215 ll.
cmn'crqml I"
we haw won
many-body C
is tedious um
Wawfum't i( u;
expansion up
particle ('nrre'
low part in] W;
‘
5M5.
("
rim-Icon furrw
arts. and now
ldllon Of 51“ (
)
lmpOrtant h.“
alldltlon U) l}
. [I
dr .
spin, and isospin channels, a linear combination of correlated Gaussians forms a
dense, non-orthogonal set. As a consequence, different sets of coefficients A (or a)
represent the wavefunction equally well and none of them are really indispensable.
This property of basis functions suggest the idea of a completely random selection
of non-linear parameters.
Furthermore, the nuclear interactions are strongly state dependent and for a realis-
tic description of a nucleus, many different orbital, spin, and isospin channels need
to be considered. The number of possible channels grows rapidly with the number
of nucleons; but again, some channels will receive larger weights in the wavefunction
than others.
Also, in principle, all basis terms could be expressed in the same Jacobi set which
would, as in three-body models, require inclusion of high partial waves to reach
converged results. The convergence in orbital momentum is very slow in general, as
we have seen for three-body results in Chapter 2, for example. Moreover, in practical
many-body calculations, matrix element computation involving high partial waves
is tedious and time consuming. On the other side, mixing different Jacobi sets in the
wavefunction delivers faster convergence and only the low terms of the partial wave
expansion are needed. This is because different Jacobi sets bring in different inter-
particle correlations and, as demonstrated by Eq. (2.27) for a three-body problem,
low partial waves in one Jacobi set may contain higher partial waves in other Jacobi
sets.
As a result, even for fairly simple systems, such as 4He bound by realistic nucleon-
nucleon forces, the combined number of orbital, spin, and isospin channels, Jacobi
sets, and non-linear parameters becomes prohibitively large. Therefore, the calcu-
lation of all of the matrix elements and diagonalization including all potentially
important basis states of type Eq. (3.2) may be out of the question. In SVM, in
addition to the random selection of the Gaussian non-linear parameters, a ran-
dom selection of the orbital, spin, and isospin channels, as well as Jacobi sets is
53
...:l
itttrmluw"
in tllf‘ “1"
this Sé‘i'tl‘ :
FLLALM'ittg t he so .
ailiiil-allfl-("rrt )1
haéié Statttfi<. 'l lit
l.GfluTHH'A.
is axst-mltlm
lallhrk n
illllHllL',
WU Rundnt
vuut int
9
.. Through our]
mnquUHI()
ne'w C'atitlitlut‘
3. Among tho 1'
provious Stt‘p
D‘ ..
Ndrtl all Ht
Ea -
—. construction
'32»?er ' -
z. 15 gamut “-j
33le i ‘
. improvwl fut
r .-" .
‘ . l‘f') Hr‘:
‘fihuw --
iElIf—S‘ l (g I \
- ,. ) ' (.
fixm‘ ,
In ”10 mm f
. 1 . ll
Same
“HV'
. as; l
introduced. Random sampling also eliminates the danger of local minima traps
in the multi-dimensional parameter space. The linear expansion parameters c in
Eq. (3.9) are determined via energy matrix diagonalization, as explained before in
this section.
Following these arguments, the wavefunction in Eq. (3.9) is constructed term by term in
a trial-and-error method. Let us assume that the wavefunction already contains N —- 1
basis states. Then the “construction” part of the optimization procedure reads:
1. Generate M random candidates to find the N th basis function. Each new candidate
is assembled in two steps:
(a) Pick a Jacobi set and an orbital, spin, and isospin channel randomly from
among all possible channels.
(b) Randomly generate non-linear coefficients A (or a) from a “physically” rele-
vant interval.
2. Through energy matrix diagonalization, calculate the energy with the N basis states
formed out of the already adopted N —— 1 states complemented with each of the M
new candidates.
3. Among the random candidates, find the one providing the lowest energy in the
previous step, add it to the wavefunction and increase the basis dimension to N.
Discard all other random candidates.
The construction process is continued as long as some minimum amount of binding
energy is gained with the acceptance of each new basis state. After that, the energy
can be improved further without increasing the basis size by fine-tuning the non-linear
coefficients. The “refinement” is done by applying steps 1(b)—3 to the already admitted
basis states, i.e. by changing the non-linear parameters of one of the basis states already
present in the wavefunction. The non-linear parameters of a single basis state are changed
in the same way as before by randomly selecting the best combination. The refinement
54
.al
‘5 done C}'(‘ll('«'*
351011;; as bitlti
wavefunction t1.
may be newh‘rl
We C()ll('lllti
was tlm't’lttpt‘tl '
calculations tn-
neétitmn-rii'h ix
version. the nut
3H9. and ‘lllc.
simple ls harmt
inter-Clustcr um
In such (alt-ulut i
spin-orbit and ('
3.2 4He i
Having the frunn
main this ninth
Eiltif‘ln’f‘ and tin
t't‘l> I
-.tfl nin-rtN-t ipi
(=-' , ‘
.at. be obtaitn'tl
with ,,
strong nip”
(TWE- v
J.i Ergpnfp 1. ) _
' L ~‘ )
3'2'1 Inter.
”11‘: {.. . .
I’JNQI)”. (1
min: '
ml important
pref! .i
Jl‘. .onc lllll‘
is done cyclically for each term in the wavefunction. Again, this fine-tuning is continued
as long as binding energy is being gained. After that, more basis terms are added to the
wavefunction through the construction algorithm. Several construction-refinement cycles
may be needed before full convergence of the binding energy is reached.
We conclude this section by outlining a microscopic cluster version of SVM which
was developed to account better for clustering in light nuclei and make the many-body
calculations more tractable [132]. Over the years, the cluster version was applied to
neutron-rich isotopes of elements with masses between six and eleven [75,77,133]. In this
version, the nuclei are comprised of individual nucleons and microscopic clusters of 3H,
3He, and 4He. The intrinsic wavefunctions of composite clusters are constructed from
simple ls harmonic-oscillator Slater determinants of a common width parameter. The
inter-cluster wavefunction is modelled by correlated Gaussians and optimized by SVM.
In such calculations, effective central nucleon-nucleon forces with occasional addition of
spin-orbit and Coulomb interactions were used to bind the nuclei under study.
3.2 4He in the stochastic variational model
Having the framework of SVM outlined, we now wish to construct the wavefunction of 4He
within this model. In Chapter 5, such 4He will serve as a core in 6He. For 4He bound by
effective and realistic nucleon-nucleon forces, SVM results are in perfect agreement with
other microscopic models. For effective nucleon-nucleon interactions, converged results
can be obtained with less than a hundred correlated Gaussians [74]; for realistic forces
with strong repulsion at short distances, several hundred basis states are needed to reach
convergence [125].
3.2. 1 Interactions
In microsc0pic calculations, the choice of the effective nucleon-nucleon interaction is of
crucial importance. If one wants a model to have anything to do with the real physical
problem, one must make sure that the inter-nucleon force is appropriate for all subsystems
55
9-:I
u awn"
w
appearing in t't
of valence neu'
theéf‘ tonhgui‘n‘
g~,’1t~n_i_9-§|
gpht’flC'dl
mut‘it'ls. ‘
torrent. 1
the nthei
3f the W-
hom? w." r .
if] Barri):
other Init
by a sing-f
printed (
:n tnuierl
other (or
in the ‘1
{titer hat
:iirznnish
irinrniati
form suit
tnnurirml
nicriels.
In .\Ii
53 must rt
mint lH]
Burris. t l 1
iiit? in mi
”Ki." thro:
will (Ting)
03mm" )tl
spherical model from [79]. There are several differences between MiCH and the other
models. The most obvious is the treatment of the valence part. In MiCH, a functionally
correct, exponentially decaying valence basis is adopted from three-body models, while
the other microscopic cluster models employ Gaussians of some sort in the valence part
of the wavefunction to reduce the computational demand. As we argued in Chapter 1,
however, a Gaussian basis may not be the Optimal choice to capture the asymptotic trends
in Borromean nuclei. Another difference is the amount of details built into the core. In
other microscopic cluster models applied to 6He, the 4He core is normally approximated
by a single Slater determinant of single-particle harmonic oscillator wavefunctions. It was
pointed out in [76,134], however, that such a simplified treatment of the core may lead
to underbinding of 6He relative to the three-body break-up threshold. It was argued that
other configurations such as 3N + N (here, N stands for a nucleon) should be present
in the 4He to account for a possible distortion of the core in 6He. In MiCH, on the
other hand, a converged 4He from Section 3.2 is used as the core in 6He, which should
diminish the underbinding problem. Last, but not least, in Chapter 5 we will extract
information about the 4He + n + n decomposition of 6He modelled within MiCH in the
form suitable for feeding calculations of transfer reactions through which halo nuclei are
commonly studied. This step has not been reported from other microscopic structure
models.
In MiCH, the wavefunction in Eq. (4.3) is assembled in two steps. First, the core
is constructed as a free nucleus. Once optimized, the core’s variational parameters are
excluded from optimizations in the second step of the wavefunction assembly. In other
words, the distortion of the core due to valence neutrons is not considered explicitly, and
like in many other microscopic models, a possible distortion of the core is accounted for
only through the core-valence antisymmetrization. 1n the second stage of the wavefunc-
tion construction, the focus is on the valence part containing discrete (nlagflval, .1301),
continuous non-linear (p0), and continuous linear variational parameters (c). These va-
64
sail
c
|
\
..‘_ 3‘ ‘1
ienre parameters
is minimizml. IIP
need a reliable at
Essentially th
tered in both the
iriuiifi' model. I [It ‘
ei'aluami. In the
raneter p” in all
gerameters until 1
signal. but hoe-am
t‘lciN‘tl form. TINT:
parameters,
In )IiC'II. the
With a functionall‘
and the valence pa
to he an easv wav
appropriate for the
one could consider t
if Gaussians and ti
all -' . '
(altttlattons ton]
in“ -
td compromise th
In lIiCH. the
only oJ
E‘l- (4.7). In tl
II.‘ ('(‘LV
Spittist' ' '
. .rspin (llllll’ll.\fll)1
Due to 0
Mr mum-
]' .
””5- HOW:
I
(.-
91 0110 runs
lence parameters are to be varied until the expectation value of the Hamiltonian:
_ (‘I’IHI‘I’>
E _ W (4.7)
is minimized. Here, ‘11 is the wavefunction from Eq. (4.3). To minimize the energy, we
need a reliable and efficient method to compute integrals in Eq. (4.7).
Essentially the same integration/optimization problem from Eq. (4.7) was encoun-
tered in both the three—body model in Chapter 2 and SVM in Chapter 3. In the three-
body model, the integrals involved in Eq. (4.7) are low-dimensional and can be easily
evaluated. In the three-body model, it is then possible to have the same non-linear pa-
rameter p0 in all valence terms and to simply enlarge the space of discrete variational
parameters until energy convergence is reached. In SVM, the integrals are multidimen-
sional, but because of the Gaussian basis, the integrals can be evaluated analytically in a
closed form. Therefore, SVM can rely on a random trial-and-error selection of variational
parameters.
In MiCH, the situation is different. The wavefunction in Eq. (4.3) combines a core
with a functionally very different valence part. Upon the action of Ame—val, the core
and the valence part of the wavefunction are blended together and there does not seem
to be an easy way to move between permuted Jacobi sets in order to find the set most
appropriate for the computation of a given matrix element. To make calculations easier,
one could consider expanding the valence hyper-spherical functions in Eq. (4.2) in terms
of Gaussians and thus make the core and the valence part functionally identical. Then,
all calculations could be carried out in the SVM fashion. However, such an expansion
could compromise the long-distance behavior of the wavefunction, and so was abandoned.
In MiCH, the only option seems to be a numerical evaluation of the matrix elements in
Eq. (4.7). In the case of 6He, the integrals involve 6 x 3 = 18 spatial and 2 x 6 = 12
spin-isospin dimensions.
Due to their numerical evaluation, matrix elements in Eq. (4.7) come with uncertain-
ties. However, one must still be able to evaluate the integrals with sufficient accuracy to
65
sail
1
new -“
perform a meanin
from trivial. esiN'V
statedependent. 3
methods of oil-grit
an of numerical ii
for example. if i ll]
1H: Simpson's r
points 15‘]. There
with N. There-ft tI't‘
tegration spar-e tn
the door to Monte
oithe integral fan.
tariational .\l( )lll ('
pmhlems of the t)"
4.2 Variat
.lc‘ stated in the I)
matrix elements Int
To simplify the not.
Eq. {4.3) depending
it = *
[11 ..... 72,1}l (i
The expectation val
t a
th. ground-state ene
0n the lefph
and sir]:
th '
(“tight-hand side t
perform a meaningful variational calculation. The computational problem at hand is far
from trivial, especially in the context of nuclear physics, where the interactions are highly
state-dependent. Moreover, due to the dimensionality of the integration space, standard
methods of on—grid integrations are out of the question. In conventional quadrature meth-
ods of numerical integration, the accuracy depends on the density of the integration mesh.
For example, if one uses a d-dimensional cubic mesh to evaluate a d—dimensional integral
using Simpson’s rule, the error scales as N "4/ d, where N is the total number of mesh
points [138]. Therefore, as the dimension d increases, the error falls off increasingly slowly
with N. Therefore, a better way to evaluate multidimensional integrals is to scan the in—
tegration space to find the regions most relevant for a given physical problem. This opens
the door to Monte Carlo integration techniques, in which the statistical error in the value
of the integral falls off as N ’1/2 regardless of the integral’s dimensionality. In particular,
variational Monte Carlo has proven to be a very powerful tool to tackle mathematical
problems of the type of Eq. (4.7).
4.2 Variational Monte Carlo
As stated in the previous section, we are dealing with a variational problem in which
matrix elements must be evaluated by means of multi-dimensional numerical integrals.
To simplify the notation in this section, we shall use ‘II(7"', s, t, p) for the wavefunction in
Eq. (4.3) depending on all spatial (F: {F1,. . . ,FA}), spin (3 2 {X1,. . . ,XAI) and isospin
(t = {T1,. . . ,TA}) degrees of freedom as well as on a set of variational parameters (p).
The expectation value of the Hamiltonian H with the wavefunction gives an estimate of
the ground-state energy:
(‘1’0‘3 8, t,p)|H|‘1’(7'“3 s,t,p)>
01’0", 8. t,p)|‘1’(fi 8, t,p)>
(H (10)) = Eh?) = (4-8)
On the left-hand side, the braces denote expectation (mean) value of the operator, on
the right-hand side, they mean integration over all 7", s and t. By minimizing E(p) with
66
respect to all variational parameters, one obtains the approximation to both the energy
and the wavefunction of the true nuclear ground state. Integrals involved in Eq. (4.8) are
to be computed by means of variational Monte Carlo (VMC).
Mathematical foundations of various Monte Carlo techniques can be found for ex—
ample in [139]. VMC has been applied to problems in nuclear [66,140] as well as other
areas of physics [138]. For a general description of VMC, we shall lean mostly on [138—140]
supplemented by other references. The term “variational Monte Carlo” comprises two as—
pects of the problem in Eq. (4.8). The “variational” part refers to the variational nature
of the problem, whereas the “Monte Carlo” part is responsible for the actual evaluation
of the integrals involved. The general concept of VMC, however, does not provide any
guidance on moving within the space of variational parameters that would help us move
towards the variational energy minimum.
In the following text, we first] focus on numerical aspects of Eq. (4.8). The Metropolis
algorithm along with other techniques used to evaluate integrals will be described. Then,
VMC will be applied to a simplified case of 6He to convince ourselves that the integrations
are done properly in MiCH. The section on VMC will conclude with the discussion of
parameter optimization techniques used in MiCH.
4.2.1 Monte Carlo essentials
The Monte Carlo approach to the evaluation of multi-dimensional integrals relies on
statistical sampling and averaging of the integrand. In this section, the essentials of the
Monte Carlo integration are presented.
Adhering to the physical problem at hand, let us consider many-body functions de-
pending only on A spatial coordinates concisely denoted as 7" 2 {F1,. . . ,FA}; complica-
tions due to spins and isospins will be discussed later. The goal is to find the integral of
some function f (7"):
12/f(fldf, (4.9)
where the integral is taken over the relevant space of 1". An underlying idea of Monte Carlo
67
integration lies in transforming the integrand into a product of two auxiliary functions
19(1") and g(F). The former function, also called the importance or sampling function, is
chosen first such that it obeys:
pm 2 o, / pm (,4: 1, (4.10)
and hence may be ascribed a meaning of a probability density. Consequently, the vector
1" may be considered a random variable. The other function g(F) is computed accordingly
as g = f / p. Then, the original integral can be written as:
I = / gm node (4.11)
and interpreted as nothing more than the mean or expectation value of the function g('F).
Due to its dependence on the random vector 7", the function 9(7‘) is a random variable
distributed around its mean value I with variance:
vartg) s / tgta - Iran at. (4.12)
In principle, the value of I may be obtained by drawing an infinite set of mutually
independent random vectors from the distribution p0") and computing the sample average
of local values g(7"):
1
I: 1' —
Ngnoo N
M2
g (7701)) . (4.13)
n=l
Here, Fm) means the n—th set of vectors F. A Monte Carlo estimate of I may be obtained
by averaging over a finite sample:
1 z [N = N E 90702)). (4.14)
68
The variance of function 9(7") from Eq. (4.12) can be estimated in an unbiassed way as:
var(g) z chug) _=_ N _ ,aittg), (4.15)
where 012V(9) 1s a commonly computed biassed estimator of the variance of values g(F (n) )
distributed around their mean value I N3
)2 (g)%§::l( [g( FM)” —IN]2. (4.16)
The quantity 0N_1(g) is called the unbiassed standard deviation of the sample g(F(n)).
In the nomenclature of quantum Monte Carlo, the vector F is said to be a walker,
wandering around the integration space. A random stop, 77(71), of the walker is called
a configuration or a sampling point, and a chain of sampling points is referred to as
a random walk. In the present work, we prefer to use the term integration point over
sampling point. Referring back to Chapter 3, an integration point F (7‘) consists of a set
ofspotsr(),...,F/(1n).
It is important to realize that different random walks may yield different values of
I N- In fact, for any N, the Monte Carlo estimator I N by itself is a random variable
distributed with its own probability density, the expectation value of which is equal to
I, and the variance of which is:
var(g)
N .
var(IN) = (4-17)
Using the Monte Carlo estimator from Eq. (4.15) for var(g), the variance of integral
estimates I N can be estimated as:
012V_1(9).
var(IN) z N
(4.18)
We point out that, when the original integral Eq. (4.9) is estimated by Eq. (4.14) on short
random walks, the distribution of values of I N may not be Gaussian and so (/var(1 N)
69
can not be attributed the typical meaning of an error bar. As N —+ 00, however, the
central limit theorem shows that the distribution of values I N will converge toward a
normal distribution regardless of the sampling distribution p(F) or the distribution of
local values g(F(")). Only then is it meaningful to estimate a one-standard-deviation
error of the Monte Carlo estimate of the mean (also called the standard error of the
mean or simply the error):
58(1N) = Elli—(91, whenN ——» oo. ' (4.19)
W
Therefore, the error bar on the integral estimate IN will decrease as 1/\/N regardless
of the integral dimensionality in Eq. (4.9). We emphasize again that the Monte Carlo
estimate of the integral IN and the error Se are trustworthy only when they are estimated
on large samples of statistically independent integration points F (n).
A judicious choice of the importance function significantly reduces the variance for
a fixed sample size. The variance of the integral estimate var(IN) in Eq. (4.17) would
vanish for a constant function g(F) = const = I . However, this choice is not available
since the integral I is not known a priori. In practice, we want an importance function
that matches the general behavior of the function g(F). In many quantum—mechanical
problems, the importance function is taken to be the square of the actual wavefunction.
Once the importance function is chosen, the integral in Eq. (4.9) can be estimated
by means of Eq. (4.14). We rely upon the MetrOpolis algorithm [141] for generating
a random walk with integration points distributed according to the sampling function
p(F). Considered a golden standard for integration space sampling, this algorithm has
been employed in nearly all other variational Monte Carlo calculations. The description
of the Metropolis algorithm with all non-trivial details can be found for example in [139].
For our purposes, we use the following simple version of the algorithm encountered in
most practical applications:
1. Given a walker at point F, generate a trial vector Ftrial randomly from within a
3A-dimensional cube of volume (Ar)3A surrounding the point F.
70
2. Calculate the probabilities p(F) and p(Ft,.,-al). The acceptance probability for a move
from F to fit-tat is given by the expression:
.. _. - 17(7vtrial)
P 'r —-> 7‘ - = min 1, ——-— .
( trzal) { 130—3) }
In actual calculations, the ratio in brackets is compared with a random number
between 0 and 1; if the ratio is greater, the proposed move is accepted.
3. If the move is accepted, set F = Fm-al and return to step 1. Otherwise, discard the
point Fm-al and generate the next trial move from the original position F.
The very first sampling point is chosen completely randomly. By construction, the al-
gorithm satisfies the condition of detailed balance. This condition ensures that if many
walkers originating from different positions are launched simultaneously, at any time
later the number of walkers flowing from one integration point to another is the same
as the number of walkers flowing in the opposite direction. Therefore, any point in the
integration space can be reached by the walker from any other point.
Despite its simplicity, the Metropolis algorithm is of a great power, as it can be used
to sample essentially any importance function regardless of the number of dimensions.
Another advantage is that to generate a walk the importance function does not need to
be normalized because the acceptance probability P(F ——> Fm-al) depends only on a ratio
of local values of the importance function. There are, however, a few complementary
disadvantages of the algorithm. First, the sampling is correct only asymptotically. The
initial integration points generated depend on the starting point and should be discarded.
In all our calculations, at least 1,000 generated integration points are discarded before
local values of the operators are first evaluated. Second, successive integration points are
correlated, which violates the assumption of their statistical independence needed to make
the Monte Carlo estimates reliable. This correlation is obvious because the new point F in
step 3 is either equal to F from step 1, or is somewhere nearby. Consequently, successive
local values of the function g(F) are likely to be correlated. Due to such correlations, the
71
effective number of independent samples is less than the actual number of points in the
random walk, which slows down the convergence of the integral estimate in Eq. (4.14)
and also makes the formulae in Eq. (4.15) and Eq. (4.19) underestimate true statistical
deviations. Therefore, great care must be taken to ensure that the integral is estimated on
a sufficiently large set of statistically independent configurations F (n), as will be discussed
in Section 4.2.2.
Having outlined the Monte Carlo integration, we now turn back to the evaluation of
expectation values of physical observables depending on the many-body wavefunctions
\II = \II(F, s, t):
(wow), (I?
(0) = W = / t (4.20)
W W).
where O is an operator and (I) S.t denotes the inner product in the spin-isospin space. In
quantum mechanics, the square of the normalized wavefunction is a good candidate for
the sampling function:
(lqunst
Pm = W (4-21)
With this choice of the importance function and the following definition of a local value
of the operator 0:
010m 5 (M11031, ’ (4.22)
Eq. (4.20) takes a very simple form:
<0) = / 010m an at (4.23)
suitable for the Monte Carlo evaluation. At each integration point, a local value in
Eq. (4.22) is calculated and the expectation value (0) and its error are estimated by
means of Eq. (4.14) and Eq. (4.19). Local values of total energy will be denoted as E106.
In practical calculations, the norm (\II]\II) of the wavefunction is often not known and
in fact is not needed. It is because the norm appears in Eq. (4.20) that we can adapt the
Monte Carlo machinery to the computation of (O) and never actually compute the norm
72
of the wavefunction. Formally, the importance function in Eq. (4.21) is normalized to
unity to make Eq. (4.23) work; however, as we explained before, the Metropolis algorithm
does not care about whether the importance function is normalized or not. Therefore, to
generate a random walk needed to evaluate Eq. (4.23), p(F) = (\It|\I!)s,t can be used.
In a more general case, any importance function having sufficient overlap with the
square of the actual wavefunction can be used to sample the physical space. One then talks
about correlated sampling, in which the expectation value of an operator is computed as:
wow, 3 .
%P(fl dr [0mm win an dr
(0) = = , (4.24)
(\I’l‘p>3,t —¢ —'
f—pfif—PW) d7“ [10(5) P(F) d7”
where w(F) is a local weight defined as:
w(F) E 0PM)” (4.25)
190"”) '
Both integrals on the right-hand side of Eq. (4.24) are approximated by finite sums of
the type in Eq. (4.14). As far as the expectation values of operators are concerned, the
sampling function in correlation sampling does not need to be normalized because its
overall normalization would enter both integrals in Eq. (4.24) and so would be cancelled
out.
The material contained in this section is sufficient to evaluate multi-dimensional inte-
grals encountered in the variational problem in Eq. (4.7). Further discussion on technical
details and the implementation of VMC can be found in Appendix A. Before we proceed
further towards techniques used to optimize variational parameters, we have to make sure
that the integral estimates are reliable. As mentioned before, the Metropolis algorithm
has several drawbacks that need to be thoroughly examined.
73
4.2.2 Can we trust ourselves?
In this section, we examine the inner workings of the numerical Monte Carlo integrations
presented in the previous section. In particular, we look at correlations between local
values inherent in the Metropolis algorithm. We discuss several methods to attenuate the
degree of such undesired correlations.
As a test case, a simple 6He is considered bound by the central part of the Minnesota
interaction from Eq. (3.11) with the mixture parameter set to its standard value u = 1.0.
This interaction is similar to MN from Section 3.2. To optimize the wavefunction of the
4He core, both K- and H-like Jacobi channels and all orbital channels with l S 2 and
L = 0 are present in the model space. Variationally Optimized 4He core contains 20
Gaussian basis terms and its binding energy is -30.77 MeV. To assemble the 6He guinea
pig, a single valence term in the T Jacobi basis with K = 2, la; = ly = L = S = 0 and
may = 0 is attached to the core as in Eq. (4.6). The non-linear parameter in the valence
part is set to p0 = 1.0 fm to reproduce approximately the experimentally known size
of 6He. The choice of the valence channel is given by its major role in the three-body
wavefunction of 6He in Section 2.4.3. Therefore, in spite of its simplicity, this 6He is
expected to provide a reliable testing ground. Unless noted otherwise, this 6He is used in
all calculations in the rest of the current section.
As mentioned in Section 4.2. 1, successive integration points generated by the Metropo-
lis algorithm may be correlated, often very strongly. Let us consider a random walk
containing N integration points. At each point, a local value of some function g(F) is
evaluated. To assess quantitatively the degree of correlation between local values g(F(”))
k: integration points apart, we use a biassed estimator of auto-correlation coefficient:
N—k
1 _. n .. n
r(g,k) = (N _ k) ”1211(9) 1; [g(rt )) _ 1N] [g(rt +10) _ [N] (4.26)
with the notation preserved from Section 4.2.1. By definition, r(g, 0) = 1, and for N = 2,
r(g, 1) = —1. The auto-correlation coefficient takes positive values for highly correlated
74
local values g(F(”)) and vanishes for uncorrelated samples. The distance I: will be called
the correlation distance.
In this section, the focus is on correlations between local values of total (kinetic plus
potential) energy, because in light of the variational principle in Eq. (4.8), the total
energy is of primary interest. Hereafter, by energy we mean the total energy, and a local
value of the total energy is shortened to a local energy, unless stated otherwise. In the
MetrOpolis algorithm, the degree of correlation can be controlled by the linear size A,—
of the 3A-dimensional cube from which trial moves are drawn. To illustrate the effect of
Ar, three random walks were produced with values of Ar = 4.5 fm, 1.4 fm and 0.15 fm,
corresponding to Metropolis acceptance probabilities (or rates) of about 5%, 50% and
95%, respectively. Each walk contains 10,000 integration points and the sampling function
was taken as the square of the wavefunction of 6He. For each walk, auto-correlation
coefficients were computed and are shown in Figure 4.1 along with fragments of Monte
Carlo histories of local energies.
It is evident from Figure 4.1 that local energies are tightly correlated, especially for
extrema of MetrOpolis acceptance rates. The explanation in terms of Ar is simple. Imag-
ine a walker at some point F in space with a presumably large probability p(F), Then a
trial move drawn from a cube with large A,— may easily end up in a region with much lower
probability, and so be rejected, which gives raise to flat sections in Monte Carlo histories.
On the other hand, when AT is small, a trial move does not disturb the probability p(F)
too much and will most likely be accepted, which results in fairly smooth Monte Carlo
histories. These arguments are also reflected by the very slow decay of auto-correlation
curves in Figure 4.1. On the other hand, the case with moderate Metropolis acceptance
produces fairly weakly correlated local energies. Based on this observation, we shall here—
after adhere to the lore of quantum Monte Carlo holding that the Metropolis algorithm
should accept about fifty percent of trial moves in order to produce good results.
Even when the Metropolis acceptance rate is close to 50%, local energies are still
correlated, but the degree of correlation is quickly attenuated as the correlation distance
increases. This observation suggests an improved sampling algorithm in which every
75
000- l 1 L l l n 1 ml "
1.0” . [Ar=1.4m
SE . .
b O
2 05' '.. ‘
LU
‘c.’
O
01
.0 .
o
o ‘250 560‘730‘1000 0‘10‘20‘30L4o‘50
integration point k
4‘;
(a) Monte Carlo histories of the first 1,000 local (b) Auto-correlation coefficients between local
energies. energies computed from all 10,000 integration
points in each random walk for different corre-
lation distances 11:.
Figure 4.1: Effects of correlations in the Metropolis algorithm on local energies. The
results are for three random walks with Ar = 4.5 fm, 1.4 fm, and 0.15 fm corresponding
to Metropolis acceptance rates of about 5%, 50% and 95%, respectively.
integration point is decorrelated a certain number of times before local values of operators
are evaluated again, as is schematically illustrated in Figure 4.2. The auto-correlation
curve in Figure 4.1(b) with A, = 1.4 fm suggests that about 5—10 decorrelation steps
should suffice to substantially decorrelate local energies. In Figure 4.3 we show auto—
correlation curves for local energies produced in three new independent walks with 1,
10, and 30 decorrelation steps. From this figure it is also clear that having about 10
decorrelation steps is indeed good enough to break correlations between local energies.
Using more decorrelation steps hardly improves the results.
An alternative approach to decorrelated sampling is the data reblocking (or bunching)
method [142]. The method works as follows. At reblocking level 0, a set of local values
{g(F(")),n = 1,... ,N} is considered a set of N (0) E N blocks, each block holding a
single local value and thus having a block average value of gill) E g(F(")). Then, for any
higher reblocking level b, new blocks are formed by merging the two neighboring blocks
76
walker at F
evaluate local values
ti
r->[ Metropolis: attempt to move to find]
v accept reject
T : Ttrial
it t
walker at F
not decorrelated yet decorrelated
Figure 4.2: Flow chart for the decorrelated Metropolis algorithm.
from the previous level. Therefore, the new block averages are equal to:
_ b 1 _ b— _ b—l
9:1,) 2 5 [gén,1)l+gén, )] b=1,2,... (4.27)
with n’ = 1,... ,Nlb), where N (b) = int(N(b‘1)/2). Here, the operation int means trun-
cation to the nearest integer. The reblocking continues as long as at least two new blocks
can be formed. At each level, the sample mean value and its error Slab) are computed by
applying Eq. (4.14) and Eq. (4.19) to the block averages (77(3). One can even estimate the
numberof
1.0]: " decorrelation
steps:
0.8 r I 1
L . O 10
j; 0.6t * 3°
m3 0.4 " ' 'l
h 0.2 - -
l
0.0 r ‘
1 L 1 . I A L A L l
o 10 20 30 4oL50
k
Figure 4.3: Auto—correlation coefficients between local energies as a function of the corre-
lation distance k in three independent walks with 1, 10, and 30 decorrelation steps. Each
walk contains 216 integration points and the Metropolis acceptance rate is about 50%.
77
error on the standard error Séb) as Séb)/ (/ 2(Nlb) — 1) [142].
The idea behind the reblocking algorithm is the following. As the reblocking level
increases, each new block contains more and more original local values and so, on the basis
of the central limit theorem, block averages approach independent Gaussian stochastic
variables. When the sample size N is a power of twol, the sample mean is invariant under
the blocking transformation. Furthermore, for statistically independent block averages,
the standard error Séb) is also blocking-invariant. For correlated data, blocking typically
yields increasingly uncertain estimates of Sigh) and the best error estimate is obtained for
the smallest blocking level beyond which Séb) saturates. At very high reblocking levels,
the error estimate may become unreliable because of the small number of remaining
blocks.
To see the effect of reblocking, in Figure 4.4 the bunching algorithm is applied to the
data from Figure 4.3. In general, energy block averages become less correlated as the
level of reblocking increases. However, for the walk with a single decorrelation step the
reduction of correlations is still fairly slow, the error estimate Séb) barely saturates, and
the error estimated directly from the local energies (reblocking level 0) severely underes-
timates the true error. In agreement with previous observations, about ten decorrelation
steps are needed for the error estimate of the mean energy to quickly form a distinct
plateau, the appearance of which is a fully convincing signal that energy block averages
have become statistically uncorrelated and saturated error estimates can be trusted. Note
that for well decorrelated walks saturated error estimates are almost identical to those
for no reblocking. In most of our calculations, blocks are formed from 100 local values
which corresponds to reblocking levels 6—7.
For the walk with 30 decorrelation steps from Figure 4.3, the evolution of the mean
energy and the energy error estimate is plotted in Figure 4.5. At the beginning of the
walk, the running mean energy is poorly defined due to low statistics. Later into the
walk, the energy curve flattens and the statistical error in the energy is reduced. Even
1Otherwise, at some reblocking levels, single excess blocks are dropped in order for Eq. (4.27) to
work, which results in a data loss.
78
.l
i
l
l
_ . numberof
0'8 n _ decorrelation
_ I ‘ steps
0.4 L ' - 0.3 ' 1 l ‘
. ' . 1 ° 1° .
00122292:868 58§: . E '30 "1)
E a P 6 * ‘ : 0.2 [— . I i i i f 0 u
" -0 4 numberof . a . ' l .
' decorrelation O 3 a: _ «
steps ' ‘ "(n 0.1 - - it -
4 I C O 9 O
.03 0.110 .. ‘.99::....993Iif5
* 3° 1:
_1.2 - l A L . 1 . 1 0.0 1 1 1 l . 1 . l . 1
0 3 6 9 12 15 O 3 6 9 12 15
b b
(a) Auto-correlation coefficients between succes— (b) Estimates of the standard error of the mean
sive energy blocks (A: = 1). energy.
Figure 4.4: Effects of bunching on data from Figure 4.3 at different reblocking levels b.
though not visible in the figure, the error falls off as N ‘1/2 towards the end of the walk
where a sufficient number of integration points has been averaged.
For an additional insight into the effects of correlations, Figure 4.6 shows block aver-
ages and unbiassed standard deviations of local values within blocks of kinetic, potential
and total energies for data from Figure 4.3. First, we notice large cancellations between
kinetic and potential energy resulting in a fairly narrow interval of block averages of total
’ I I I I I 4 I I I I WI
- L -
26 3t .
g . 1
a -27 ~ ‘ a 2 P r
a a . .
LU -28 h - m0 1 _ L -t
-29 ~ ] o - -
0‘150‘3001450‘600 0‘150‘300 450.600
block block
(a) Mean energy computed by applying (b) Energy error estimate computed by apply-
Eq. (4.14) to block averages. ing Eq. (4.19) to block averages.
Figure 4.5: Mean energy and energy error estimate computed along the walk with 30
decorrelation steps from Figure 4.3. Local energies are divided into blocks of 100 values.
79
200
0
lb
0
o
N
O
O
.
- ‘.- f{'o'
- -,'.-...-: £33153“ 4.44.4 @4444 .44, 4.- ti
- ' .0. :0
I 33?? "fin’dt 9.0L." '
block average [MeV]
O
-200
200
0
-200 lo A 1 A l 1 l A l A 4 l
O 150 300 450 600 0 150 300A 450 600
block block
(a) Block averages. (b) Standard deviations within blocks.
Figure 4.6: Block values of kinetic (solid black squares), potential (empty red circles),
and total (solid blue stars) energy for data from Figure 4.3. Each block contains 100 local
values. Panels from bottom to top correspond to walks with 1, 10, and 30 decorrelation
steps.
energy. The undesired spread of block averages and deviations within blocks is reduced
as the number of decorrelation steps increases, which can be explained as follows. Cor-
related local values tend to follow one another more closely (see cases with A,» = 4.5 fm
and 0.15 fm in Figure 4.1(a)) such that there will be entire blocks of them with a fairly
small internal spread, but which lie on average quite far from the walk average. On the
contrary, well decorrelated local values are more evenly distributed on both sides of the
walk average, the consequences of which are a more eflicient averaging within blocks re-
sulting in a narrower distribution of block averages and larger deviations within blocks.
Again, taking more than ten decorrelation steps hardly improves the results.
In summary it is obvious that the Metropolis algorithm used to generate random
walks may easily provide biassed results. In this section, the focus was on the energy as
the observable crucial for variational optimization, but the observations made are valid
80
for other observables as well. Based on the arguments presented, in this work we adhere
to the following rules:
a The Metropolis acceptance rate should be close to 50% Of proposed trial moves.
0 Each walk is decorrelated by a sufficient number Of decorrelation steps. In most
calculations, we use 30 or more decorrelation steps.
0 The reblocking algorithm is used to reliably estimate the error in the energy.
0 For increased accuracy, results computed on several independent random walks may
be averaged.
Appendix B contains additional reliability tests.
4.2.3 Wavefunction optimization
Having introduced the Monte Carlo background needed for evaluation of integrals in
Eq. (4.7), we now turn to the actual problem of wavefunction optimization. The goal is
to optimize variational parameters in the valence part Of the wavefunction in Eq. (4.3).
In this section, these variational parameters are called simply parameters. VMC, as pre-
sented so far in Section 4.2, provides a general framework to compute expectation values
of energy for a given set of parameters, but the method must be accompanied by addi-
tional tools to give any sort Of guidance in the parameter space.
From a physics standpoint, the parameter optimization is achieved via minimization
Of energy in Eq. (4.7). From a practical standpoint, the variational problem is more
complicated due to the statistical evaluation of matrix elements required. Local energies
fluctuate around the estimated mean energy and, as we have seen in Figure 4.5, along the
course Of a random walk the running mean energy may drop below its converged value.
For a. given set of parameters, it is a matter of luck whether the energy estimate is below
or above the true energy, which is not known anyway. Therefore, energy estimates may
be misleading, especially when one compares energies for difl'erent sets of parameters. It
has been argued [143,144] that a numerically more stable parameter Optimization can
81
be achieved by minimizing the variance, Eq. (4.16), of local energies, or by minimizing
a linear combination of the energy and the variance of the energy [144]. The major
argument for variance minimization is that at the minimum the variance is known to
be zero a priori. Moreover, each term in the sum Eq. (4.16) is bounded from below
by zero. The major drawback of variance minimization is that, from the physical point
of view, one typically seeks the lowest energy, which is not guaranteed to be delivered
by variance minimization unless the true eigenstate is found. Furthermore, it has been
Observed [145] that energy Optimized wavefunctions give on average better expectation
values for other observables. Sometimes, the energy and variance minimizations are used
to Optimize different parts of the wavefunction [146]. In the present work, the expectation
value of energy is minimized. As the energy is minimized, the variance Of local energies
is reduced as well.
Regardless of the actual Optimization procedure, one needs to ensure that the energy
is really being lowered. Correlated sampling introduced in Section 4.2.1 is a useful trick
commonly employed to roughly disentangle the effect Of a small change in the parameters
on energy from ambiguities arising due to the statistical sampling. One begins with a
wavefunction \Ilref containing reference values Of parameters. For this wavefunction, a
reference random walk {F ("),n = 1,. . . ,N} is generated and energy Eref is estimated
from Eq. (4.23). Suppose (some of) the parameters in ‘I’ref are slightly2 disturbed result-
ing in a new wavefunction \Il. Then, instead Of generating a new walk for ‘11, we can use
the reference walk to estimate the energy E corresponding to \II. Because both energies E
and Eref are computed on the same walk, statistical ambiguities potentially arising from
different walks are suppressed. One normally looks at the difference between the two en-
ergies because they are highly correlated, and the error of the difference is much smaller
than errors on the two energies themselves. Using Eq. (4.24) with p(F) = (‘I’refl‘I'reflat3
2The term “slightly” needs tO be defined with care. It includes not only a slight change of continuous
parameters, but also a change of discrete parameters due to a change, addition and/ or removal Of valence
channels. A quantitative measure Of “slightness” will be provided later in terms of local weights.
3The importance function for correlated sampling does not need to be normalized, as discussed in
Section 4.2.1.
82
for E, the energy difference is estimated as:
N
1 Eloc(F(n))w(F(n)) 1 -o(n)
_, _ — (Eloc)re (7A ) (4'28)
N
AE=E—Eref% Zn:
with local values at any point F (7‘) defined as in Eq. (4.22) and Eq. (4.25):
(lelwist
(414) , (Elam):(Wrele'q’mW w: (414)..
s,t
(Wreflq’refisJ , (wreflq’ref)s,t.
(4.29)
Eloc Z
Taking advantage Of correlated sampling, the same reference walk can be used in
several subsequent adjustments of parameters before a new reference walk should be
generated for the best new wavefunction. A good reason to produce a new reference walk
is when the energy has been lowered significantly, i.e. when the absolute value of the
energy gain AE between the reference and the best new energy becomes larger than
the error on the reference energy, ideally by at least a factor of two. Another reason to
update the reference walk is if the wavefunction \II with adjusted parameters starts to
differ significantly from ‘I’ref: which is signalled by local weights. Correlated sampling is
reliable provided that local weights w('r}',) in Eq. (4.29) do not significantly exceed their
average value. If the parameters are changed too much, a single weight or a few large
weights will dominate over the others in Eq. (4.28), thus biasing the energy difference
estimate. TO avoid this negative effect in MiCH, all local weights should be smaller than
about 10-20 times their average.
The valence part Of the wavefunction in Eq. (4.3) is a linear combination Of basis
functions, each Of which depends on continuous and discrete parameters. When these
parameters are changed, the linear expansion coefficients 0 in Eq. (4.3) can be determined
via the energy matrix diagonalization in Eq. (2.25). Formally, the wavefunction is written
as \11 = Z,- c,-\II,; and matrix elements H i j and I 23' in Eq. (2.26) are computed in correlated
sampling on a reference walk generated by the previous best guess for the wavefunction.
Again, as for expectation values Of Operators in Eq. (4.24), the overall normalization
of the sampling function is not needed. Unlike the overlap matrix I, the Hamiltonian
83
matrix H estimated on a finite random walk may not be symmetric. Following the
arguments in [146], we do not symmetrize the energy matrix. The overlap matrix may
be ill-conditioned because Of the possible linear dependence Of valence terms. When that
happens, the overlap matrix is regularized by a singular value decomposition from Sect.
2.6 in [147] and the non-symmetric generalized eigenvalue problem of type Eq. (2.25) is
solved with the aid Of numerical libraries [148].
In general, non—linear parameters p0 do not need to be the same in all valence terms in
Eq. (4.3), as is common in three—body models. In fact, as suggested by the success of SVM,
it could be beneficial to mix valence terms with different values of po, eSpecially because
the core-valence antisymmetrizer in Eq. (4.3) removes the orthogonality properties Of
valence terms. We have tested several methods to Optimize non-linear parameters in tan-
dem with energy matrix diagonalization and correlated sampling. A stochastic selection
in the fashion of SVM turns out to be inefficient because of the computational demand
required to calculate the energy for a single set of parameters. There exist determinis-
tic Optimization methods analyzing the local dependence of the mean energy (Newton
method) [149] and the wavefunction [150] on variational parameters. These deterministic
methods have been tuned for and proved efficient in atomic and molecular physics, but
upon testing them in MiCH, they do not seem to be adequately robust to meet our needs.
In MiCH, the added complexity is most likely due to spin-isospin contaminations in the
wavefunction and highly state-dependent, (non-)central, nuclear interactions. Also, at
the beginning of the Optimization route, the nucleus is three—body unbound. This can
be seen, for example, from Figure 4.5, where 6He containing a single valence term is
bound by about -27.5 MeV, a value to be compared with the binding energy of the 4He
core, -30.77 MeV, given in Section 4.2.2. All the Optimization methods mentioned tend
to break the nucleus apart, unless the radius of the nucleus is constrained. Additionally,
correlated sampling is reliable only for small changes in non-linear parameters.
The easiest way tO control the size of the nucleus is to make the non-linear parameter
p0 the same in all valence terms, and that is the approach in this work. Even then,
however, this parameter is a true variational parameter, which needs to be tuned to
84
minimize the binding energy. From a distant view, our Optimization method resembles
that of three-body models: for an Optimum value of the parameter p0, the number of
valence channels in the wavefunction is increased until convergence in energy is reached.
However, on closer inspection, there are some major differences.
All the previous comments on parameter Optimization are valid in general. We now
turn to 6He, the nucleus to be studied in Chapter 5. First, we consider the nucleus bound
by a soft-core effective central nucleon-nucleon interaction, such as MN in Section 3.2.1.
In this case, the 4He core contains only basis terms with L = S = 0, as we have seen in
Section 3.2.2. This makes valence neutrons in spin-singlet and spin-triplet states orthog-
onal. In the three-body analysis of 6He in Section 2.4.3, spin-triplet states were present
in the wavefunction only due to the spin-orbit interaction. It is then suflicient to con-
sider spin-singlet valence terms only. The optimization begins with a single, K = 0 or 2,
mag = 0 valence term attached to the core. Valence terms with higher hyper-momenta
and degree Of hyper-radial polynomials are added to the wavefunction until convergence
in the binding energy is reached. The value of po is adjusted to keep the rms proton radius
Of 6He close to its experimental value. To avoid high partial waves in the valence part,
both Y and T Jacobi configurations are mixed. After each addition Of (few) new valence
terms, linear coefficients c in Eq. (4.6) are determined via energy matrix diagonalization.
Despite numerical evaluation Of energy and overlap matrices, the lowest eigensolution of
Eq. (2.25) is numerically stable even when the matrices are of appreciable size, Of the
order of 100 x 100 elements. Starting with a converged wavefunction, different values
of po are tested using correlated sampling and energy matrix diagonalization to finally
locate the energy minimum.
To prevent numerical difficulties, valence terms with may 75 0 should be avoided until
a preliminary convergence of mag = 0 terms with the hyper-momentum K has been
reached. The reason is that higher-order hyper-radial terms may produce extraordinarily
large local values of kinetic energy making the Monte Carlo energy averaging harder to
converge. In the present work, the problem of large local kinetic energies is called the
problem of bad points and is discussed in detail in Section BI.
85
Parameter Optimization is more involved in the presence of the non-central spin—
orbit force between nucleons. As mentioned in Section 3.2.2, the spin-orbit force mixes
L = S = 0 and L =2 S = 1 Gaussians in the wavefunction of 4He. When attached to
such a core, valence spin-singlet and spin-triplet terms are not necessarily orthogonal in
the fully antisymmetrized wavefunction of 6He; some of them may be almost orthogo-
nal, though. Consequently, the overlap matrix I may contain many very small elements
which are hard to distinguish from statistical noise. The energy matrix H is affected
less severely. Under such circumstances, the energy matrix diagonalization may be nu-
merically unstable yielding unreliable eigenvectors of linear coeflicients c. Thankfully,
the lowest energy eigenvalues are still numerically stable. These are effects at the edge
Of numerical stability and given the statistical sampling, their severity varies between
random walks. To circumvent this problem, a major modification was introduced into
the parameter Optimization procedure described above for 6He bound by central forces,
namely a comparative optimization on two independent random walks. The details Of
this improved method are rather technical and as such are relegated to Appendix C.
86
Chapter 5
6He in MiCH
We are now ready to put the model developed in this work to the test. In previous
chapters, we have described all components going into the model including numerical
techniques needed to evaluate matrix elements and the variational optimization method.
In this chapter, the model is applied to the ground state of the simplest two—neutron halo
nucleus, 6He.
The chapter starts with a study Of antisymmetrization effects in 6He. As we will
see, these effects are crucial for binding Of 6He. Then, basic observables computed for
optimized 6He wavefunctions will be presented and compared with experimental data and
values Obtained in other theoretical models. To appreciate the amount of details going into
different models, the discussion will continue with a more detailed comparison of results
Obtained within MiCH and within the three-body model from Section 2.4. Finally, the
chapter will be concluded by an application of MiCH to a two-neutron transfer reaction
involving 6He.
5.1 Antisymmetrization effects in 6He
In this section, the effects of antisymmetrization Operators in Eq. (4.6) are studied. Unlike
in three-body models, wavefunctions in MiCH can be properly antisymmetrized to ac-
count for the fermionic nature Of nucleons. Core-valence antisymmetrization should make
87
states occupied inside 4He unavailable for valence particles. The antisymmetrization Of
the valence part ensures the Pauli principle between valence neutrons, a requirement
important especially in the Y Jacobi basis in which, unlike in the T basis, the valence
basis terms from Eq. (4.2) do not meet the Pauli principle by construction.
In the current section, “a (valence) channel” means 6He containing the MN-SO
4He core from Section 3.2 and a single valence term characterized by a set of numbers
(K, lx,ly,L, S, nlag,Y/T}. Ground states of both 4He and 6He have J7r = 0+, and so
the total angular momentum and parity of each valence channel must be J30) = 0+. The
system is bound by the MN-SO interaction from Section 3.2.1.
We focus on valence channels with the lowest hyper-momenta, namely with K = 0
and 2.1. Possible combinations Of angular momentum quantum numbers for such channels
are shown in Table 5.1. The table also shows whether a given channel in the T Jacobi
basis is blocked by the Pauli principle between valence neutrons. Due to their trivial
(constant) hyper-angular dependence, K = 0 channels contain the “lowest” core-valence
s-waves (lm = 0) in the Y Jacobi basis, and as such are expected to be the most core-
valence Pauli-blocked.2 On the other hand, K = 2 channels are expected to be crucial
for the structure Of 6He, as has been demonstrated in Section 2.4.3. For K = 2 channels,
squares Of the Raynal—Revai coefficients for angular transformations between Y and T
Jacobi sets from Eq. (2.27) are shown in Table 5.2.
Figure 5.1 shows binding energies of channels with K = 0 and 2, and mag = 0 and
1 for different values of the valence hyper-radial parameter p0. For each channel, three
energy curves are shown differing by whether the antisymmetrizers AMI and Acme-val
in Eq. (4.6) are active or not. For each channel, both antisymmetrizers are first disabled,
then Aval is switched on, followed by the activation of floofe—va'l . We observe in Figure 5.1
that when the core-valence antisymmetrizer does not act, the valence antisymmetrizer
has barely any effect on T Jacobi channels. It must be so because these channels satisfy
1K = 1 channels are not allowed because they would have to combine 3- and p-waves due to Eq. (2.8)
The parity of such channels would be negative which violates the requirement of positive parity of the
valence part.
2Remember that in a first approximation 4He can be thought Of as four nucleons sitting in the lowest
s—shell.
88
Table 5.1: Possible combinations of angular momentum quantum numbers for K = 0 and
K = 2 channels.
Pauli blocked by Eq. (2.23)
K I“: I” L S in T Jacobi basis 81138
0 0 0 0 0 no K = 0 s-waves
2 0 0 0 0 no K = 2 s-waves (spin-singlet)
2 1 1 0 0 yes K = 2 p-waves spin-singlet
2 1 1 1 1 no K = 2 p-waves spin-triplet
the Pauli principle between valence neutrons by construction.
For a given "lag, K = 0 s-waves are essentially identical in Y and T Jacobi bases,
as we discussed in Section 2.2. Without core-valence permutations, K = 0 s-waves with
may = 0 and 1 form deep energy minima in Figure 5.1(a). These minima, however, are
removed upon the action of Acme—val. Interestingly enough, the fully antisymmetrized
K = 0, may = 1 channel is bound more than the most trivial Of all valence channels, the
channel with K = 0, may = 0, which is simply a manifestation Of the Pauli principle.
The K = 0, mag = 0 channel puts a neutron in the Y basis into the radially most trivial
s—wave (la; 2 0) motion around the core, which makes this channel strongly forbidden by
the core-valence Pauli blocking. On the other hand, due to its non-trivial hyper-radial
dependence, the channel with K = 0,7146,g = 1 contains “less” trivial core-valence s—waves,
which makes it “less” forbidden by the Pauli principle.
We start the analysis of K = 2 channels by K = 2 s-waves in the T Jacobi basis,
i.e. by Figure 5.1(b). When the valence neutrons are not antisymmetrized with nucleons
Table 5.2: Squares Of the Raynal-Revai coefficients (lglgllzly) KL from Eq. (2.27) for
angular transformations of K = 2 valence channels between Jacobi sets. (lglgllxly) K L
corresponds to a transformation from the unprimed to the primed Jacobi set, or schemat-
ically unprimed —+ primed.
[(00|00)2ol2 [<11|00)2012 [(000020]2 [<11l11)20l2 Illlllllml2
Y —> T 0.04 0.96 0.96 0.04 1.00
T —> Y 0.04 0.96 0.96 0.04 1.00
89
20 20 n
g 10 _ €13 " ~
E E. '
A A 12 — -
O o
I o I .
L L
LU UJ 8 - -
3 -1o « 3 l
I I 4 _
“L «a, 4
m LU [
-2 A l A l A l A l A l A l A #1 4 L444 1 A l A l A
Q12 0.4 0.6 0.8 1.0 1.2 1.4 1.6 (23.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
po[fm1 no lfml
(a) K = 0 s—waves in T basis (b) K = 2 s-waves in T basis
20 ' I ' I I I I I 20 I I I I I I
l ’ 4
5' 16 — - S‘ 16 - -
o o
E. ’ l 2. ' J
’3 12 [- - 3 12 t-
:I: 1 :1: .
L ‘L,
m a 4 « m 8 - —
7:” l i 3:3 i
“L 4 h “2., 4 b q
in » 1 LL!
%.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 00.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Po 11ml Po [fml
(c) K = 2 s-waves in Y basis ((1) K = 2 p—waves spin-singlet in Y basis
20 m 1 v 1 .
S‘ 16 — a
o
2 i
A 12 - .4
O _
I “'89 - O
‘L, —-— OFF, OFF
U1 8 -o—OFF. ON a
,L -x—o~. on
o “'39 = 1
«i 4 - -- OFF. OFF ‘
Lu - 0- OFF. ON
-'°ON'ON 1.41141;
%.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
no [fml
(e) K = 2 p—waves spin-triplet in T basis
Figure 5.1: Three-body binding energy of valence channels as a function of the non-
linear parameter p0 for different antisymmetrization settings. The legend is the same in
all panels. For each channel, the OFF/ ON switches indicate whether the corresponding
antisymmetrizer in Eq. (4.6) is active or not: the first switch controls Aval, the second
switch is for Acme—val. Energy curves are constructed in correlated sampling on walks
for reference values of po; for each curve, reference energy is depicted by a star. Error
bars appear on reference values only.
90
inside the core, 6He behaves as a pure three-body system core + n + n. It is three-body
unbound with binding energy monotonically heading towards the three-body threshold
(i.e. towards the binding energy of 4He) as the hyper—radial scaling length p0 increases.
Larger p0 implies larger average hyper-radii and thus larger 6He through the three-body
relationship in Eq. (2.29). Therefore, the gain in binding energy with increased p0 is
misleading, because the nucleus gradually breaks apart into the core and two neutrons.
This undesired trend is changed dramatically once Acorewal is switched on, because
the core-valence exchange effects deliver extra binding to the system. Such a simple,
fully antisymmetrized 6He with mag = 0 remains three—body unbound, but the binding
energy saturates for a fairly large interval of po in Figure 5.1(b). This saturation forms a
foundation on which a variationally optimized 6He in Section 5.2 will eventually become
bound against the three-body break-up. Using Table 5.2, K = 2 s-waves in the T Jacobi
basis consist of 96% K = 2 p-waves spin-singlet state in the Y basis. Therefore, in the
Y basis a neutron is mostly in a relative p—wave motion around the core, making the
channel K = 2 s—waves in the T Jacobi basis mostly Pauli allowed.
The response of the other two K = 2 spin-singlet channels from Table 5.1, s-waves
and p—waves spin-singlet in the Y basis, can also be easily understood. We first look at
K = 2 s-waves in the Y basis, i.e. by Figure 5.1(a). In the absence of both AMI and
Acme—val antisymmetrizers, this channel (for a given may) is bound the most among
all K = 2 channels, as can be seen from Figure 5.1. The reason for this is that in this
channel one of the neutrons would be in an s-wave (l3 = O) motion relative to the core,
making the entire channel fairly bound. The binding is weaker than for K = 0 s-waves
though. As soon as the valence particles are antisymmetrized, however, several MeV of
the binding energy in Figure 5.1(c) are lost in this channel, which can be explained as
follows. Using Table 5.2, we can write schematically:
K=2s—wavesinY —> 4%K=2s—wavesinT+
96% K = 2 p — waves spin — singlet in T.
91
However, the p—wave admixture in the T basis is completely eliminated by the action of
AMI, because it is Pauli blocked in Table 5.1, and the original channel effectively turns
to K = 2 s-waves in the T basis. Upon a rotation back to the Y basis and using Table 5.2
again, one obtains:
val
K=2s—wavesinY3—4——>K=2s—wavesinT———>
4% K = 2 s —- waves in Y + 96% K = 2 p — waves spin — singlet in Y.
The result is a dramatic structural change: the original s-waves in the Y basis become
mostly p-waves in the same basis putting a neutron into a p—wave (lx = 1) relative to
the core. The net result is the above-mentioned loss of binding. Similar analysis can be
done for K = 2 p-waves spin-singlet in the Y basis (Figure 5.1(d)):
val
K=2p—wavesspin—singletinY —A—>K=28——wavesinT——+
4% K = 2 s — waves in Y + 96% K = 2 p — waves spin — singlet in Y,
and so the structural change due to the action of Ava! is far less dramatic because the
channel remains mostly p-waves in the Y basis. Once valence particles in K = 2 spin-
singlet states in the Y basis are antisymmetrized, these channels become equivalent to
K = 2 s-waves in the T basis. Therefore, the core-valence exchange effects in these
channels can be understood on the merit of the discussion for K = 2 s-waves in the T
basis.
The remaining K = 2 channels in Table 5.1 contain valence particles in a spin-triplet
state. For a given mag, there is only one such state in the Y basis and one in the T
basis. Therefore, they must be essentially identical upon rotations of Jacobi bases, as is
also demonstrated by corresponding Raynal-Revai coefficients in Table 5.2. The energy
gain in these channels due to antisymmetrization effects is not large enough to produce
a saturated energy curve in Figure 5.1(e).
92
In summary, at the level of single valence channels, the antisymmetrization exchange
effects have significant impact on 6He. Not only do they enforce the fermionic nature of
the nucleus, but they also deliver extra binding to the system. Similar analysis could be
done for valence channels with higher hyper—momenta, but we have limited our discussion
to valence channels that are crucial for the structure of 6He. We convinced ourselves that
general observations made in this section depend neither on the value of the mixture
parameter u in the Minnesota interaction nor on inclusion of the spin-orbit force.
5.2 Converged 6He
Having understood the behavior of single valence channels, we now proceed to the op-
timization of the wavefunction of 6He. Results are presented for two cases: MN and
MN-SO defined in Section 3.2.1. More emphasis is put on MN—SO because it employs
a more realistic nucleon-nucleon interaction due to the spin-orbit force. In both cases,
the mixture parameter u in the central part of the interaction was adjusted to bind 6He
by about the right amount against the break-up into 4He and two neutrons. Essentially,
the interaction mixture parameter is the only free parameter in MiCH. The Coulomb
interaction is neglected since it would barely shift absolute binding energies of both 4He
and 6He by about the same amount. The wavefunctions for the 4He core are taken from
Section 3.2.2.
The convergence of the binding energy of 6He relative to the three-body threshold
with the number of valence terms in the wavefunction is shown in Figure 5.2. The varia-
tional optimization technique used in MiCH was described in Section 4.2.3. Here, a few
comments are given on chronology in Figure 5.2.
The case MN is discussed first. The construction of the wavefunction begins with a
single K = 0, ”lug = 0 valence channel with which the nucleus is three-body unbound.
The scaling length p0 is set to 0.80 fm to keep the rms proton radius of 6He close to its
experimental value 1.91 fm. Next, all valence channels with K g 10 and mag = 0 in both
Y and T Jacobi bases are gradually added to form a fairly well defined wavefunction
93
20, 'O ' ‘ ' - MN.poadjusted
5.1% . o MN-SO,p0=O.45Ofm
Q fir ,5
a 2' :
. » i
I 1_ f, .
L '1 th 1h hold
u.l . §§ roe-bod); res ‘
a O
I
“L
LIJ
‘6
. exp. $08.-
1 0%‘350 11.03;” . 4‘
I
‘
l A 1 A l A 1 l
o 25 50 75 ‘100‘125
# of valence terms
Figure 5.2: Convergence of the three-body binding energy of 6He with the number of
valence terms included in the wavefunction. In MN, the variational parameter p0 is
adjusted along the optimization route; in MN-SO, the results are for a fixed value p0 =
0.45 fm. See text for details. Error bars were not computed for all points, and even when
present, they may be smaller than the actual symbol.
before valence terms with higher hyper-radial orders may are considered. The parameter
p0 is slightly enlarged because the radius has become smaller due to stronger binding.
The three-body binding energy of a still three-body unbound 6He is about +0.5 MeV
in Figure 5.2. Next, all valence terms with K = 0 and 2 and may = 1 are added to
the wavefunction and the three-body break-up threshold is finally crossed. By crossing
the three-body threshold, the binding energy as a function of po forms a variational
minimum. Horn this moment on, p0 is adjusted to approximately minimize the energy.
It then takes another 118 valence terms to reach the converged value —0.90 MeV for the
three-body binding energy. At the end, all spin-singlet valence terms with K g 12 and
may 3 5 in both Y and T bases are included in the wavefunction.
Figure 5.3 shows the dependence of the three-body binding energy and the rms pro—
ton radius on the scaling length p0 for converged 6He. In the MN case, the variational
energy minimum is located around p0 = 0.45 fm. The steep reduction in binding below
p0 = 0.40 frn in Figure 5.3(a) reflects trends observed in Figure 5.1. By reducing p0, the
nucleus becomes smaller in Figure 5.3(b) because the valence neutrons are forced to stay
closer to the core. As a consequence, the binding of 6He is reduced due to the increase
94
'0.7 I I _l_ MN 1 _85 I fl f ' I ‘ I I I
—o— MN-SO
% -O.8 " " -—1
2. E. 1.80 -
3 8
V3: -0.9 ’ ‘01:, 1
“J Q
,L "'A 1.75 ‘ ‘
f -1.0 - 1 «1‘70-
0v
m l f
1 L 1 1 1 . 1 . 1 1 1 4 ._ I . 1 _._ MN
'1 '1 0.3 0.4 0.5 0.6 0.7 1'70 0.3 0.4 0.5 0.6 0.7
Po [fml Po [fml
(a) Three-body binding energy. (b) Rms proton radius.
Figure 5.3: Dependence of the three—body binding energy and the rms proton radius of
converged 6He on the non-linear parameter p0. All curves are constructed in correlated
sampling on walks for reference values of po. Reference values of observables are depicted
by stars.
of total kinetic energy of the system. On the other side, for very large values of po, the
valence neutrons are forced to spend more time farther from the core, and the size of the
nucleus increases and the binding becomes weaker.
The optimization procedure in the MN-SO case is more complicated, as described
in Appendix C. In this case, the wavefunction is tailored to a specific value of po. The
wavefunction is first optimized for p0 = 0.70 fm, a value that prevents 6He from get-
ting too large at the beginning of the optimization when the nucleus is still three-body
unbound. Using the method of a comparative optimization outlined in Appendix C, a
fairly converged wavefunction is constructed. For this wavefunction, a reference walk is
produced. On this walk in correlated sampling, the binding energy is estimated for an
auxiliary 6He containing a set of all 222 valence terms with K g 14 and may 3 5 for
different values of po. The corresponding energy curve is shown in Figure 5.3(a). Be-
cause for this auxiliary 6He only energy eigenvalues are reliably determined in the energy
matrix diagonalization, neither error bars nor radii are shown. The energy minimum is
formed around p0 = 0.45 frn, a value for which the wavefunction is optimized again, and
the convergence plot is shown in Figure 5.2. At the point where 6He becomes three-body
95
bound, a subset of about 15 valence terms with K S 6 and may 3 2 is included in
the wavefunction. Altogether ninety carefully selected valence terms having K S 14 and
may 3 5 are needed to reach energy convergence in the MN-SO case.
Several extra remarks regarding the variational Optimization are appropriate at this
point. First, the convergence plots in Figure 5.2 are not unique. If the history plots were
constructed again, they would look differently depending on several factors, such as the
order in which valence terms are added to the wavefunction; the exact values of po used
throughout the optimization; statistical effects due to random sampling etc. However,
we convinced ourselves that the converged results from Figure 5.2 would be reproduced.
Second, in Chapter 2, the valence basis in Eq. (2.18) is orthogonal and complete, and
converged results obtained within the three-body model should be independent of po.
In MiCH, the orthogonality of valence terms is destroyed by the core-valence antisym-
metrizer. As expected, the non-linear parameter p0 is then a variational parameter, as
would also be suggested by a formation of energy minima in Figure 5.3(a). Last, in the
following discussion, converged results for 6He are for p0 = 0.45 fm in both MN and
MN—SO cases.
Once the wavefunctions have been optimized, we can calculate binding energies and
rms point matter, point proton and point neutron radii for 6He. Radii are computed as
square roots of expectation values of operators in Eq. (2.28), Eq. (2.30) and Eq. (2.32).
The results are shown in Table 5.3 along with experimental values and results obtained
in a variety of other models: the macroscopic three-body model from Chapter 2 and
Section 2.4 in particular, SVM [76] as a representative of microscopic cluster theories,
and the Green’s function Monte Carlo (GFMC) [117] representing microscopic models.
Preliminary results from the MN model have been published in [153].
Let us first comment on experimental values of radii. What has become experimentally
known with a great accuracy are nuclear charge radii (1:3)” 2 of 4He and 6He. However,
in our calculations, nucleons are treated as point particles, and rms point proton radii
(7%)”2 are calculated. Following [47], the charge and the point proton radii are related
96
Table 5.3: Binding energies E and three-body binding energies E3b0dy in [MeV] and rms
point nucleon radii in [fm] of 4H6 and 6H6 from various models along with experimental
values. MN and MN-SO are results of this work, the other models are the three-body
model from Chapter 2, SVM [76] and GFMC [117]. Experimental proton radii were com-
puted by means of Eq. (5.1) using charge radii from references cited in the table. Values
labelled with * were computed; for radii, the relationship from Eq. (2.33) was used,
GFMC three-body binding energy was obtained using binding energy -29.4(1) MeV of
6He from [117]. The thickness of the neutron halo is defined as Ar = (1%)”2 — (7%)1/2.
MN MN-SO 3body SVM GFMC exp.
4He E -30.85 -3093 N/A -25.60 -28.37(3) -28.30[118]
(1,2,)1/2 1.40 1.40 1.40 1.41 145(0) 146(1) [151]
E31,“), 090(5) 102(3) -0.98 -O.96 -1.03(10)* -097[25]
(1%,)1/2 241(1) 232(1) 2.49 2.42 2.55(1)* 248(3) [152]
233(4) [18]
6He (1%)”? 181(1) 175(1) 1.86 1.31 191(1) 1.91(2)[47]
(7,2,)1/2 267(1) 256(1) 2.75 2.68 282(1) 2.72(4)*
2.51(6)*
Ar 086(1) 081(1) 0.89 0.87 091(1) 081(4)
060(6)
as
N
(2%) = 02> — (12%) — <63.) (51)
E,
where 02%,)” 2 = 0.895(18) fin [154] is the rms chargeradius of the proton, (Ra) =
—0.120(5) frn2 [155,156] is the mean-square charge radius of the neutron, and N and
Z are nuclear neutron and proton numbers. We used values 1.681(4) fm [151] and
2.054(14) fm [47] for charge radii of 4He and 6He, respectively. The corresponding proton
radii are shown in Table 5.3. In literature, matter radius of 6He has been extracted from
relevant interaction cross-section data; however, the extracted values disagree, as listed
in Table 5.3, according to the type of analysis performed.
In our calculations, a free 4H6 in Table 5.3 turns out to be overbound and smaller
relative to experimental data. The fact that we differ from experimentally known abso-
lute binding energies was expected because of the effective nucleon—nucleon interactions
employed. We are, however, mostly interested in three-body—like features of 6He. The
97
three-body binding energy is approximately reproduced after a slight tuning of the mix-
ture parameter u in the Minnesota interaction. Our proton radii of 6He are smaller
than they should be which may be a consequence of a smaller 4He core. On the other
hand, matter radii are comparable with those deduced from experiments. Perhaps due
to stronger three-body binding, MN—SO 6He is slightly smaller than its MN counterpart.
Nevertheless, proton and neutron radii change consistently so that the thickness of the
neutron halo does not change dramatically between MN and MN-SO.
Next, MN and MN—SO results are compared with those of other models listed in
Table 5.3. In the three-body calculations in Section 2.4, the binding energy and the size
of 4He core do not enter the actual three-body calculations, but the radius of the core
is needed to compute the size of 6He. In the three-body picture, radii of 6He are simply
related to the radius of the core via Eq. (2.29) and Eq. (2.31). For the best comparison
between our and three—body results, we assume the same radius of 4He in the three-body
model as in MN-SO. Then, in a naive three-body picture, the larger size of 6H6 in the
three-body model (when compared to MN-SO) is solely due to the valence neutrons living
on average slightly farther from the core. The three-body model is also useful to assess
how strongly radii of 6He depend on the size of the core. If the radius of the core is
increased to its experimental value 1.46 fm, then radii of 6He become (7%,)1/2 = 2.51 fm,
(7%)”2 = 1.90 fm, (7,2,)1/2 = 2.77 fin and Ar = 0.86 fm. With these new values, the ex-
perimental proton radius of 6He is perfectly reproduced, and the neutron halo shrinks a
bit.
Within SVM, 6He has been studied in the past repeatedly [76,132,133]. In Table 5.3,
SVM results obtained in [76] are quoted where central and spin—orbit Minnesota and
Coulomb interactions were employed. In the later reference, several different cluster com-
positions were considered to study break-up of the core in 6He. In Table 5.3, results for
4He correspond to model 02 in [76], i.e. to an a-particle wavefunction of which is a su-
perposition of three 1s harmonic oscillator Slater determinants with common oscillator
parameters set to minimize the a’s ground-state energy. Due to this very simple picture,
SVM 4H6 in table Table 5.3 is bound less than 4He in MN and MN-SO cases. The SVM
98
results for 6He in Table 5.3 are those from model (b) in [76]. In that model, 6He was
modelled as a combination of 4He + n + n and t + t with tritons again built from simple
1s harmonic oscillators. The importance of the triton channel was first observed in [134],
where this channel was introduced to overcome the insufficient three-body binding of 6He.
SVM three-body binding energies and radii of 6He are comparable with ours, especially
with the MN model.
For the sake of completeness, we also show microscopic GF MC results in Table 5.3.
These were obtained using realistic two-body AV18 and three-body IL2 interactions. We
show GFMC results to point out that by using modern realistic potentials in microscopic
calculations, absolute binding energies and proton radii of 4He and 6tHe can indeed be
reproduced. However, as we argued in Chapter 1, questions may arise about how micro-
scopic models treat asymptotic regions so important for Borromean halo nuclei.
We have also computed the point nucleon density distributions. For the more realistic
MN-SO case, they are plotted in Figure 5.4 along with density distributions obtained in
other models from Table 5.3. For comparison, the figure also contains the proton (equal
to neutron) density for the MN-SO 4He. These nucleon densities are calculated as simple
6—function expectation values:
2::
1 1 q, A 1472,25 4 .. q, 52
471—7‘2 (why) :1 2 (T- [Tl—TONISI) ( ' )
with :l: for proton and neutron densities, respectively. Here, 7’; is the operator of isospin
projection from Appendix A. For a nucleus with mass number A, the integral is carried
over all spatial coordinates 1",- as well as all nucleonic spins and isospins. To compute
the densities corresponding to the three—body model, we constructed an auxiliary wave-
function of 6He of type Eq. (4.6) by combining the MN-SO 4He and the valence part
taken from the three-body wavefunction of 6He obtained in Section 2.4. In such auxiliary
wavefunction, the core-valence antisymmetrizer in Eq. (4.6) was switched off because the
Pauli principle was approximately taken into account when the three-body wavefunction
was constructed in Section 2.4. In logarithmic scale, proton and neutron densities from
99
density [fm'a]
r [fm]
Figure 5.4: Point proton and point neutron density distributions in 6He for models from
Table 5.3 except MN. The GFMC densities are from [66]. Also, for comparison, the proton
(=neutron) density of the MN-SO 4He is shown. All proton (neutron) distributions are
normalized to the number of protons (neutrons).
different models of 6H6 are close to one another with small differences reflecting slightly
different radii and wavefunction compositions. All models reproduce the most pronounced
property, namely the neutron distribution extending far beyond that of protons. In other
words, they reproduce the neutron halo of 6He. Depleted at short distances, the proton
density of 6H6 stretches farther out than that for a free 4He. A partial explanation of
this effect comes from the three-body model: in 6He, the a core does not sit at the cen-
ter of mass of the entire system, and its motion relative to the center of mass spreads
out the proton distribution. Due to the same effect, the neutron density in 6He is also
expected to be depleted at small distances relative to that of a free 4He, as is also visible
in Figure 5.4.
The effects of the Pauli blocking from Section 5.1 were tested further in the MN model.
First we consider a case with the core—valence antisymmetrizer switched off. When the
K = 0 valence channels are present in the wavefunction, the nucleus is three-body over-
bound by several tens of MeV, as one would expect based on Figure 5.1(a), where K = 0
s—waves are deeply three-body bound. When, however, K = 0 s-waves are all removed
from the wavefunction, the nucleus becomes three-body unbound regardless of inclusion
of valence terms with higher hyper-momenta. When the core-valence antisymmetrizer
100
is switched back on, the converged 6He is three-body bound by about -0.9 MeV from
Table 5.3. When all the lowest K = 0 hyper-spherical channels are now removed from
the converged wavefunction, the three-body binding decreases to about -0.75 MeV. It
becomes evident that, to produce a meaningful 6H8, it is not sufficient to simply neglect
the most Pauli-blocked K = 0 valence channels. Rather, all contributing valence channels
should be included in the model space and subjected carefully to the antisymmetriza—
tion. This message is important especially for few-body models in which the core-valence
forbidden states are removed approximately by different Pauli blocking techniques.
We can see that all models mentioned in Table 5.3, although different in their nature,
are in a fair agreement on most commonly computed properties of 6He. The agreement
between densities in the three-body and MN-SO models is especially remarkable given
how those densities were obtained. In MN-SO, the valence part of the microscopic wave-
function was built from the very beginning on top of a microscopic 4He core; in the
three-body model, the internal structure of the core was roughly accounted for through
Pauli blocking, and only the final optimized three-body wavefunction was attached to
a microscopic MN-SO core in order to obtain nucleon densities. In linear scale, densi-
ties from MN-SO and the three-body model are shown again in Figure 5.5. It looks like
highly integrated properties, such as radii and densities, may not appreciate the amount
of details built into different models of 6He. We may then pose a question: is it really
worthwhile to pursue a time-consuming microscopic approach to 6He if a computation-
ally cheap three—body route works so well? Provided the three-body picture of 6He is well
suited for a problem at hand, can we not simply use three-body wavefunctions for all our
needs? One way to shed more light at this issue is to compare three-body wavefunctions
directly with their analogs extracted from MiCH, with overlap functions.
5.3 Overlap functions
For a more detailed comparison with results obtained in the three-body model, the three-
body—like core + n + n information needs to be extracted from MiCH. To access this
101
r [fm]
Figure 5.5: Point proton and point neutron density distributions in 6He for MN-SO and
three—body models from Table 5.3.
information, an overlap integral between a microscopically described two-neutron halo
nucleus (\II) and its core ((1)) is computed:
A
1M Judi—1 = (2) (‘I’Jgoremjcm Tom MTcml‘I’JWMJTMT) (5-3)
The binomial factor accounts for the number of combinations to pick two out of A nucle-
ons. The integration is done over all degrees of freedom in the core (I), and so the overlap
integral I depends on degrees of freedom of two valence neutrons remaining outside the
core. The integral has a good isospin and isospin projection 1 and -1, respectively, but
it does not have a good angular momentum. It can be expanded in a complete set of
hyper-harmonics with good angular momentum Jgal from Eq. (4.1):
J A! 3 /4
I _ = J , [ a ]
[VJ val1 1 Z CJCOTeA/[Jcore JvalAI-I 07001 leral (,0) (#1H2) 7’0“! Jam! 1_1 ,
,7 l J1r val
va val
(5.4)
where C are Clebsch—Gordan coefficients. As in Chapter 2, 7w) = {K , l$,ly,L,S,T}.
The expansion is carried out in the T Jacobi basis where the hyper-harmonics W sat-
isfy the Pauli principle by construction. The numerical factor (111112)3/4 is included to
make the spatial part of hyper-harmonics orthonormal with respect to the weight factor
102
»-;1|
‘3/2 sin2600326 from the hyper-spherical volume element in Figure 2.1.3 Note
(#142)
that the hyper-radial part in Eq. (5.4) is not expanded in the Laguerre basis from
Eq. (2.16) because the basis functions ’R’r‘lag do not have a good physical meaning.
Instead, the overlap functions (9 are computed directly from:
A
’ _ 3/4
07%, J30, (P) - (2) (Ml/12) X (5-5)
5(p — p’)
< [(DJgore Tcore NITCO’I‘e ® W’Yval Jgall_1] '17,.le TMT __p5 \I’J’IrA/[J T MT ,
where the integration is carried over degrees of freedom of all nucleons. Using these
overlap functions, the three-body—like core + n + n component of the wavefunction ‘II
can be written as:
I
337143me = Z (mm/4x (5.6)
Jg0T€70a1 J30]
(I)
l J3me Tcore MTcore ® 0700’ ijral gym” mall—1 J 7f M J T MT
in the form similar to the three-body decomposition in Eq. (2.21). Moreover, the overlap
wavefunction ‘Iloverlap satisfies a three-body Schréidinger equation with an additional
source term due to the residual interaction between valence particles and those in the
core [157]. Therefore, at least in the asymptotical regions, the three-body wavefunction
from Eq. (2.21) and the overlap wavefunction \Ilo’vefmp from Eq. (5.6) should behave
similarly. On this merit, the three-body results and those from MiCH for 6He can now
be compared at the level of wavefunctions rather than integrated observables. A valence
term characterized by a set firm) will be refereed to as an overlap or a three-body channel.
For 6He, the three-body decomposition takes a simple product form:
overlap 6 _ 3 4
\II0+01_1 ( He) _ (2/3) / ¢0+00 Z 0711010 Wfi’vaz 0+1—1 (5'7)
71ml
3Even though not mentioned, this factor was also included in three-body terms in Eq. (2.18).
103
In analogy with the three-body model and Eq. (2.11), it may be useful to work with
modified overlap functions 11:
24p) = pm 00) (5.8)
Because of the orthonormality of hyper-harmonics, the norm of an overlap channel is
given by:
00 oo
_. 2 5 _ 2
SIUGlJflal —A 070711 J1r (p)p dp——‘/0 117%le (p) dp. (5.9)
val val
The last quantity is often called a spectroscopic factor. We will use this term also in con-
nection with the three-body results for 6He, where a spectroscopic factor gives the prob-
ability of a given channel in the three-body wavefunction, as we have seen in Table 2.1.
To compute overlap functions in a meaningful way, wavefunctions of both 4He and 6He
need to be normalized. The normalization of the the core’s wavefunction is known from
SVM. The norm of 6He wavefunction can be computed numerically, as is demonstrated
in Appendix D. Here, both wavefunctions are assumed to be normalized to unity. The
integration space in Eq. (5.5) is sampled by the square of the wavefunction ‘1! (6He).
We extracted overlap functions for the MN-SO 6He. To ensure small statistical errors,
overlap functions were computed on two separate random walks, each containing four
million integration points. At the end, the overlap functions from such two walks were
averaged in each overlap channel to improve statistics even further.
Ordered by spectroscopic factors, the five strongest overlap channels in the MN-
SO 6He are listed in Table 5.4. These are the only channels that could be resolved, all
other potential overlap channels have spectrosc0pic factors too small and as such are
buried in numerical noise.4 The table also contains spectroscopic factors for 6He studied
in the three-body model from Table 2.1. Surprisingly enough, not only the dominant
channels are exactly the same in MN-SO and the three-body model, but also their order
is preserved. In the three-body model, these five channels account for more than 98%
of the wavefunction. Therefore, we expect that these channels should also grasp most
4It seems that an overlap channel can be resolved if its spectrosc0pic factor is larger than about 0.01,
or equivalently about one hundredth of that for the strongest K = 2 s-waves channel. -
104
Table 5.4: Spectroscopic factors of the five dominant overlap channels in 6He. Three-body
results are from Table 2.1. Numbers in parentheses are relative errors.
T Jacobi basis
channel S
alias K lg; lg L S 3body MN-SO MN-SO / 3body
K = 2 s-waves 2 0 0 0 0 0.8089 1.1155 (0.5%) 1.38
K = 2 p-waves 2 1 1 1 1 0.1103 0.1859 (0.7%) 1.69
K = 0 s—waves 0 0 0 0 0 0.0417 0.0555 (2.1%) 1.33
K = 6 d-waves 6 2 2 0 0 0.0164 0.0266 (3.5%) 1.62
K = 6 f-waves 6 3 3 1 1 0.0078 0.0122 (3.0%) 1.56
)3: 0.9851 1.3957
of the 4He + n + n decomposition of the MN—SO 6He. In the three-body model, the
wavefunction is normalized to unity, and as a consequence, all spectroscopic factors are
less than one. In MN—SO, however, K = 2 s—waves channel has a spectroscopic factor
larger than one. To understand this difference, we need to dig deeper.
To see more clearly the difference between the three—body and the microscopically de-
rived overlap wavefunctions, Figure 5.6 and Figure 5.7 show the hyper-radial dependence
of the five channels from Table 5.4. We chose to show these functions as u(p) because of
their simpler asymptotic fall-off in Eq. (2.15). It is satisfying to observe that for all five
channels, three-body and overlap hyper-radial functions agree on their rough properties,
such as overall shape, and number of nodes. There are, however, a few obvious differences.
First, absolute values at peaks are larger for overlap functions than for three—body func-
tions. This difference is responsible for larger spectroscopic factors in Table 5.4 associated
with overlap functions. Second, overlap functions tend to peak and reach nodes (other
than the trivial node at p = 0 fm) slightly before three-body functions do. Also, peaks
of overlap functions are steeper. Third, at larger hyper-radii, overlap functions decay a
bit faster, as can also be seen from both Figure 5.6 and Figure 5.7. Overall, the overlap
functions on average put more weight on smaller hyper-radii. We could then speculate
that this shift of preferred hyper-radii might be indirectly reflected by smaller radii of
the MN-SO 6He in Table 5.3 when compared to radii from the three-body model.
105
p [I‘m] 9 [ml
(c) K = 0 s-waves (d) K = 6 d-waves
I ' I ' I '
(e) K = 6 f-waves
Figure 5.6: Hyper-radial dependence of overlap and three-body wavefunctions for channels
from Table 5.4. The three—body functions are from Figure 2.13. The legend is the same
in all panels.
At hyper-radii beyond about 12 fm, MN-SO overlap functions in Figure 5.6 and
Figure 5.7 become unreliable and their statistical fluctuations take over. The reason is
that very large hyper-radii would place two neutrons into regions very distant from the
106
100 r r I f I I
10'1r :
{‘102r 1 9“
.5. E
._. a ._.
$10 5 '5 :0-
2 ‘5
104g 1 _
t
45’ A . . . - . . .
1° 0 4 8 12 16
WIN]
(a) K=2s-waves
100: T ' F
10"]: 1,
gm 10¢ I I ‘31-
'E E I E
E 10"" i . —-
a f ’a
3 ' 1 5’
_ 104:? 1: _
-6 . . . . . - 1 . . '
1° 0 4 8 12 16
plfml plfm]
(c) K = 0 s-waves (d) K = 6 d-waves
0
10 . ' MN-SO
0 ' 4 8 A 112‘ 16
plfml
(e) K = 6 f-waves
Figure 5.7: Hyper-radial dependence of absolute values of overlap and three-body wave-
functions for channels from Table 5.4. The three-body functions are from Figure 2.13.
The legend is the same in all panels.
core. In extreme configurations, a hyper-radius of 15 fm would correspond to a di-neutron
at distance of about 13 fm from the center of the 4He core, or to two neutrons on Op-
107
|u(p)| [fm"”1
-5’ . . A .
1° 0 4 8
96m]
(a) K = 2 s-waves
12
116
-1
IU(p)I [Inn ’21
81'»
—L
o
15.1
. "um, . .
H—c
.5
o
L
A l
_L
o
in
0 ‘ 4 8
Plfml
(c) K = 0 s-waves
12
MPH [Ml/6
Iu1p)l 1mm]
10°, . . - 4
10"} 1
1021 1
10°?! [1. a
104;; l 1
105;: l 1
C
'6 L A i A l A l A .1
1° 0 4 8 12 16
plfml
(b) K =2 p—waves
10°_ . . . . . . . 1 a
10"] 1
1021 1
a.
10 gr 3
10°“,r ]
105% 1
6’ - 1 - . . 1 . . ‘
1° 0 4 8 12 16
plfml
(d) K = 6 d-waves
12‘16
p [fml
(e) K = 6 f-waves
Figure 5.7: Hyper-radial dependence of absolute values of overlap and three-body wave—
functions for channels from Table 5.4. The three—body functions are from Figure 2.13.
The legend is the same in all panels.
core. In extreme configurations, a hyper-radius of 15 fm would correspond to a di-neutron
at distance of about 13 fm from the center of the 4He core, or to two neutrons on op-
posite sides of the core at a mutual distance of about 21 fm. And because the sampling
probability is proportional to (@(GHe)|‘II(6He)) 3,15, such extreme spatial configurations
are very unlikely to be visited by a walker during the course of a random walk. Moreover,
samples in such distant regions may be highly correlated, an effect that has been seen in
other Monte Carlo calculations, for example [90].
It is interesting to look at asymptotics of overlap functions a bit closer. As two neu-
trons are pulled out of 6He, effects of the core-valence antisymmetrizer in Eq. (4.6) grad-
ually vanish, and the full six-body wavefunction decouples into its 4He + n + n asymp-
totical form. In asymptotic regions, the six—body wavefunction can be written as nothing
else but a sum of overlap functions. Therefore, the asymptotics of overlap functions goes
hand in hand with asymptotics of the many-body wavefunction in the core + n + n
cluster channel. Both three-body and overlap hyper-radial functions should fall off ex-
ponentially with the decay parameter r; depending on the three-body binding energy
via Eq. (2.13). For E3body (6He) = —1 MeV, we get n z 0.22 fm‘l. However, as dis-
cussed before, we found the value p0 = 0.45 fm to be optimal for the valence part of the
fully antisymmetrized wavefunction. Relating the two decay parameters as r. = 1/(2p0),
p0 = 0.45 fm would correspond to n = 1.11 fm‘l, or E3body z —25 MeV. In other
words, individual valence terms in our microscopic wavefunction decay much faster than
the expected asymptotic form of overlap (and three-body) functions. The asymptotical
form should be most clearly realized for K = 0 s-waves thanks to a small, yet non—
vanishing, centrifugal barrier in Eq. (2.12). In Figure 5.8, we plot again the three-body
and overlap hyper-radial functions for K = 0 s—waves along with two asymptotical forms
corresponding to decay scales n = 0.22 fm‘1 and p0 = 0.45 fm. It is gratifying to see
that the overlap function is almost perfect asymptotically in the computationally safe
region of p S, 12 fm. The right asymptotical trend is recovered regardless of the fact that
all valence terms in the wavefunction are asymptotically wrong. These results have been
obtained from the MN-SO wavefunction containing hyper-radial Laguerre polynomials
of order five and less. Perhaps, if needed, the asymptotics could be improved further
with inclusion of Laguerre polynomials of higher orders. In either case, presented results
108
4 . MN-SO 3
1 O 3body a
- " - oxp(-xp), K=0.22 fm
''''' exp(-O.5p/po), po=0.45 fm .
10'5 . -
4 6 8101211416
Pflml
Figure 5.8: Behavior of K = 0 s—waves overlap and three—body functions from Figure 5.6
at large hyper-radii. Also shown are two different predictions for asymptotics, one corre—
sponding to the three-body binding energy E3body = —1 MeV (14 = 0.22 fm‘l) of 6He,
the other to the asymptotics of individual valence terms in the MN-SO wavefunction of
6He (p0 = 0.45 fm).
clearly demonstrate that the three-body hyper—spherical/hyper-radial basis is suitable
not only for few-body, but also for microscopic calculations. The basis is flexible enough
to catch simultaneously short- as well as long-distance correlations, a quality especially
appreciable in the realm of Borromean halo systems. Based on these arguments, we be-
lieve that microsc0pic wavefunctions in MiCH have asymptotics very close to the proper
one, indeed.
Using the analogy between overlap functions and three-body wavefunctions, we can
also estimate relative probabilities of MN-SO overlap channels in the 4He + n + 11 de—
composition of 6He. Because of the orthogonality of overlap channels, we can define such
probability simply as a ratio of a spectroscopic factor and the value 1.3957, the sum of
spectroscopic factors from the MN-SO model in Table 5.4. These probabilities are listed
in Table 5.5 and they are very similar to the weights of corresponding channels in the
three-body wavefunction of 6He. Such a comparison is only approximate because the
five overlap channels account for only 98.5% of the three-body wavefunction and overlap
channels with spectroscopic factors smaller than about 0.01 were not extracted from the
MN-SO 6He. Nevertheless, mixing of overlap channels in both models is about the same.
109
v-3]
Table 5.5: Probabilities of overlap channels from Table 5.4 in the 4He + n + n decom-
position of 6H6. Three-body probabilities are those from Table 2.1. MN-SO probabilities
were computed as ratios of MN-SO spectroscopic factors from Table 5.4 and the value
1.3957 also from that table.
probability [%]
3body MN-SO
K = 2 s—waves 80.89 79.92
K = 2 p—waves 11.03 13.32
K = 0 s-waves 4.17 3.98
K = 6 d-waves 1.64 1.91
K = 6 f-waves 0.78 0.87
channel
We now present two more comments regarding spectroscopic factors. Even though not
mentioned so far, we also computed overlap functions from the MN model. Because the
central nucleon-nucleon interaction does not mix valence spin-singlets and spin-triplets in
the full microscopic wavefunction, overlap channels containing spin-triplets were absent
in the MN model. At the same time, spectroscopic factors of the spin—singlet channels
were about the same as those for the MN-SO model in Table 5.4. By not having spin-
triplet overlap channels in the MN case, about 14% of spectroscopic strength resolved in
the MN -SO case would be missing in MN. This potentially missing spectroscopic strength
is estimated as a sum of MN-SO probabilitias of spin-triplet channels in Table 5.5 and
it represents one possible way to estimate the importance of non-central forces on the
structure of 6He.
Another comment regards the fact that the MN-SO spectros00pic factor for the dom-
inant K = 2 s-waves overlap channel is larger than one in Table 5.4. This is expected [89]
because in the overlap integral in Eq. (5.3), the a particle does not sit at the center
of mass of 6He. In other words, spectroscopic factors larger than one are there due to
recoil effects. In [89], an upper limit on the spectroscopic factor of the dominant K = 2
s—waves channel was estimated to be about 25/ 16215625 times the probability of this
state in the shell-model. If we further assume that the probability in the shell-model can
be approximated by the three-body model (as was in fact done in the cited work), then
110
using Table 5.4 the upper limit on the spectroscopic factor of K = 2 s—waves channel
would be 1.26, a value indeed larger than that extracted the MN-SO model. But the
ratio of MN-SO and three-body spectroscopic factors in Table 5.4 varies between chan-
nels. This observation suggests that to account for microscopic information missing in the
three-body model, it may not be sufficient to simply renormalize the entire three-body
wavefunction by a common factor such as 25/16 suggested in [89].
Finally, overlap functions (also three—body wavefunctions) can be used to shed more
light on clustering in 6He. Additional insight is gained by calculating the probability of
finding definite distances within the three-body decomposition of 6He from Eq. (5.7):
2
P(Tn—mrcore—nn) = ng—nrgore—nn/ (2/3)3/4 Z 0712a10+g7val0+1_1 dag; (19y,
7val .
(5.10)
where rn_n and rme_nn are the valence neutron-neutron separation and the distance
between centers of masses of the core and the valence neutron pair, respectively. These
distances would correspond to lengths of vectors :31 and :52 in the T Jacobi basis in
Figure 2.1. The probability plot for the MN-SO 6He is presented in Figure 5.9. The
figure exhibits two peaks: a di—neutron-like peak positioned at about Tn-” :2 1.93 fm and
rem-3-7m = 2.63 fm (p = 3.33 fm) with the two neutrons close together located outside
4He, and a cigar-like peak at rm." = 3.82 fm and rme_.nn = 1.03 fm (p = 2.95 fm)
with the two neutrons positioned on opposite sides of the core. Qualitatively the same
clustering picture would be obtained within the three-body model and has also been
reported from other models, such as SVM [133].
The occurrence of two prominent clustering peaks in Figure 5.9 is not surprising
because the overlap channels are dominated in the T Jacobi basis by K = 2 s—waves.
Using definitions from Figure 2.1, the distances man and reme_nn in 6He are related to
hyper-spherical coordinates as
m-.. = \/2 p sin 0; r6044", —_— ,/3/4p cos 0. (5.11)
111
0.15 — -
3 0.10
0.05
probabill
Figure 5.9: Correlation density plot for the ground state of the MN-SO 6He. The di-
neutron and cigar-like configurations are shown schmetically.
Neglecting for a while the constant factors in Eq. (5.11), the hyper-angle serves as the
polar angle in Figure 5.9, and the directional distribution of the clustering probability is
given by the probability to find different hyper-angles. For K = 2 s-waves, the hyper-
angular probability is shown in Figure 5.10. There are indeed two peaks in Figure 5.10
allowing the formation of two prominent peaks in Figure 5.9. Despite the symmetricity
of the hyper—angular distribution in Figure 5.10, the clustering probability is not sym-
metrical due to mass factors in Eq. (5.11) and the influence of overlap channels other
than K = 2 s-waves in the three—body decomposition of 6He.
In summary, it becomes obvious that to appreciate the amount of details involved
in few-body and microscopic models of 6He, one should look beyond the few commonly
studied features such as the three—body binding energy, nuclear radii and nucleon densi-
ties. At the level of wavefunctions, the differences between macroscopic and microscopic
models are clear. To our knowledge, the overlap functions for 6He have been extracted
in this work for the first time from a microscopic structure model in a form that allows
their direct comparison with few-body wavefunctions. By comparing our micrscopically
derived overlap functions with three-body wavefunctions we concluded that a simple
renormalization of three-body wavefunctions may not be sufficient to account properly
for the microscopic information missing in few-body models. Besides their usefulness for
a comparative study of different structure models, the overlap functions presented in this
112
0.16'v'I-I-I-v
0.12- .
E I
EODB- .
9 1
9
O.
004- ~
0.00““L-1-‘I
0 15 30 45 60 75 90
0 [degrees]
Figure 5.10: The probability to find different hyper-angles in the K = 2 s-waves
overlap channel. For this channel, the hyper-angular probability is proportional to
P11/2’1/2(cos 26) sin26cos20 shown in the plot. The factor sin20cos20 is the hyper-
angular part of the hyper-spherical volume element in Figure 2.1.
section provide a crucial input to reaction calculations involving 6He, in particular, to
two-neutron transfer reaction models.
5.4 Two-neutron transfer reactions
An immediate practical application of overlap functions obtained in the previous section is
in reaction calculations involving 6He. In particular, overlap functions enter directly in the
formulation of two-nucleon transfer reactions. Several experiments have been performed
to study the two—neutron transfer from 6He onto hydrogen, helium, carbon, and copper
targets [53,158-161]. Of those experiments, we concentrate on p(6He,4He)t with incident
energy of 6He 25 MeV/ A [160], which is simpler to model due to the trivial structure
of the target nucleus. This reaction was later reanalyzed in [89], where several possible
drawbacks of the original analysis in [160] were indicated. It is clear from previous works
that, for a given beam energy, the reaction mechanism contains both sequential and
simultaneous transfers. In the former reaction mechanism, two neutrons are transferred
one by one whereas in the later case they are transferred both at once as a pair. However,
given that the sequential process involves the continuum and we do not yet have a fully
113
microscopic description in that case, the comparative study here will focus only on the
simultaneous transfer component, which is the only two-neutron transfer mode considered
in previous works [89,160].
Let us first analyze the transfer reaction p(6He,4He)t qualitatively. At the beam
energy 25 MeV / A, the reaction happens well above the Coulomb barrier between proton
and 6H8. Therefore, the neutron transfer can happen at any impact parameter. If the
two neutrons are transfered in a single step as a pair, it is reasonable to expect that they
are preferably picked from the di-neutron configuration in 6He. Located almost entirely
outside of 4He, the di-neutron peak in Figure 5.9 reaches its maximum at a distance
of about 2.6 fm between 4He and the di-neutron. Given the radius 1.46 fm of 4He, we
then expect the reaction to be mostly peripheral relative to 4He with small scattering
angles in the reaction center-of-mass system. In terms of hyper-radii, peripherality of the
reaction may be misleading, because the location of the di-neutron peak in Figure 5.9
corresponds to a fairly small hyper-radius of about 3.3 fm in Figure 5.6. Therefore, the
transfer should be sensitive not only to distant tails of overlap functions in Figure 5.6,
but also to their volume parts, or in other words to spectroscopic factors.
Quantitatively, the transfer reaction is analyzed within the distorted wave Born ap-
proximation (DWBA). Here, we briefly present the main ingredients of DWBA linking
us to overlap functions; details can be found elsewhere [162]. Under the 1-step DWBA,
the reaction amplitude for the simultaneous transfer of two neutrons in the p(6He,4He)t
reaction can be written in prior form as (see for example Eq. (9) in [89]):
D . — _ +
6514-4th" 4.2-44) <52
where git and ¢6He are overlaps between pairs of initial and final composite systems (t,p)
(+) (-)
6He— 4He—t
6He—p and exit 4He—t channels. The interaction causing the transition from the initial
and (4He, 6He), respectively. X p and X are the distorted waves in the entrance
to the final state has two parts. The first part is the sum of potentials binding the two
transferred neutrons to a proton: Vp_(nn) = Vp‘ml + Vp_n2. The other term 6V, the so
114
called remnant potential, contains potentials for a proton interacting with 4He and 6He:
5v = V4He-P — UGHe‘P.
Tfaditionally, both overlaps <1), and ¢6He would be taken as three-body wavefunctions
of a triton and 6He, an approach perfectly justified for the triton but arguable for 6He.
It is here where overlap functions from Section 5.3 obtained within MiCH enter the
game: they are exactly ¢6He' To examine the impact of the differences seen in Figure 5.6
between three-body wavefunctions and overlap functions on a physical observable, we
calculate the reaction cross section for simultaneous two—neutron transfer between 6He
and a proton. Within DWBA and apart from additional constants, the differential cross-
section is proportional to the square of the reaction amplitude from Eq. (5.12):
2
TDWBA (5.13)
pfior
d0
56 0‘ |
All DWBA reaction calculations presented here are finite-range and have been per-
formed with the code Fresco [163]. The triton three-body wavefunction (zit and the binding
potential V4He_p as well as Optical potentials U 6H8“) and U 4 He“ are the same as in [89].
For consistency, the nucleon-nucleon interactions in Vp“(nn) are the same as those used
to bind the triton [115]. For the optical potential between triton and 4He in the exit
channel, we adopted the parameter set I from Table I in [89], in which the potential was
fitted to elastic scattering data, thus significantly reducing uncertainties in cross-sections.
We have found the effects of the remnant potential to be large. Therefore, all presented
cross-sections were obtained with the full complex remnant term included.
The cross-sections of simultaneous two-neutron transfer in the reaction p(6He,4He)t
are finally shown in Figure 5.11. We considered two scenarios differing by treatment of
the overlap ¢6Hez three-body, where the overlap was taken as the three-body wavefunc-
tion from Section 2.4, and MN-SO with the overlap replaced by microscopically founded
overlap functions from the MN-SO model of 6He in Section 5.3. We have found that
the cross-sections are mostly sensitive to the components of ¢6He containing s-waves be-
tween the two neutrons, and between 4He and the di-neutron. This sensitivity is expected
115
10‘: ' r . I ' 6 ' ' ' I .—
, ———3body _ 3body
F4}. _--- MN-SO --—- MN-SO
1. ..... abodyx25/16 5"---._. ------ 3bodyx25/16
F: 10° : —.:~
on : a)
2. . 3 1
g. g. .
1
'0 1
10*: .
0 30 60 90 120 150 180 00 10 20 30 4o 50
0 [degrees] 0 [degrees]
(a) Logarithmic scale. Points are experimental data (b) Linear scale for small angles.
from [160].
Figure 5.11: Cross-section of the p(6He,4He)t reaction at 25 MeV/ A. Lines are theoretical
results for the simultaneous two—neutron transfer process in the three-body, MN-SO and
rescaled three-body model. See text for details.
because the reaction favors low momentum transfer. From the structure point of view,
the major role of overlap channels containing s-waves is due to their dominance in the
three-body 4He + n + n decomposition of 6He, as we have seen in Section 5.3.
At small angles, the ratio of the three-body and MN-SO cross-sections in Figure 5.11
is very close to the ratio of spectroscopic factors for K = 2 s-waves in Table 5.4, i.e.
close to 1.38. This observation reflects our suspicion that the reaction at small angles is
sensitive to almost the entire overlap functions from Figure 5.6. except perhaps at very
small hyper-radii. Moreover, at small angles, the effects due to slightly different shapes
of three-body and overlap functions are “integrated out”, and the only thing that seems
to matter is the difference in spectroscopic factors. For comparison, Figure 5.11 also
contains the three-body cross-section renormalized by an additional spectroscopic factor
25/ 16 = 1.5625 suggested in [89]. When one uses three-body wavefunctions in place of
overlaps ¢6He’ a simple additional renormalization of cross-sections is the only way to
account for missing microsc0pic structure input. Observing the similar shapes of cross-
sections in the three-body and MN-SO cases in Figure 5.11, one could try to argue in favor
of such ad-hoc renormalization of three-body cross-sections. However, we have to realize
116
that here we study only one reaction mechanism at a single energy; the cross-sections for
other reactions and / or at different energies could be more sensitive to differences between
three-body wavefunctions and microscopically derived overlap functions. Also, the three-
body renormalization factor 25 / 16 is just an upper estimate based on a simple shell-model
picture of 6H6. In either case, by using microsc0pically derived overlap functions, the
cross-section is increased by about 40% compared to that obtained by using three-body
wavefunctions, which by itself is significant given the quality of the experimental data in
Figure 5.11.
Looking at Figure 5.11, one can see the disagreement between experimental data and
theoretical calculations despite using microscopically derived overlap functions. It has
been concluded in [89] that the disagreement most likely indicates the influence of other
reaction channels not included in calculations, such as sequential transfer and/or 6He
break-up. Also, in the range of angles where the theory predicts a strong rise of the
cross-section, there are no experimental data to guide the theory in the right direction.
117
Chapter 6
Summary and outlook
6.1 Summary
Halo nuclei are composite systems with prominent features of few-body correlations. The
best examples of nuclear halo species are known to exist among light neutron-rich nuclei,
in which a single or few neutrons may be partially decoupled from the rest of the system,
from the core. The weak attraction to the core experienced by halo neutrons allows them
to swim in distant, classically forbidden regions.
Particularly interesting are two—neutron halo nuclei, such as 6He and 11Li, with two
correlated neutrons forming the halo. Typically, these nuclei are studied within few-
body models, in which the long-distance inter-cluster motion is treated properly, but
the inert—core picture used in such models is undoubtedly a simplification to the many-
body problem. Nevertheless, few-body models supply the structure information for many
reaction calculations involving two—neutron halo nuclei. On the other hand, microscopic
models find halo species very challenging and these models may fail to capture the few-
body long-distance correlations so important for halo nuclei.
At the heart of this dissertation is MiCH, a microscopic cluster model of two-neutron
halo nuclei. Designing MiCH, the goal was to combine advantages of few-body and mi-
croscopic nuclear structure models to create a microsc0pic model capable to deal simul-
taneously with short- and long-range effects in two-neutron halo nuclei. To accomplish
118
this goal, a properly antisymmetrized wavefunction in MiCH consists of a microscopic
core—like piece and a three-body-like valence part expressed in terms of hyper-spherical
functions. In the present work, MiCH has been applied to the ground state of the sim-
plest two—neutron halo nucleus, 6He, bound by the effective nucleon-nucleon Minnesota
interaction. The results for this nucleus can be summarized as follows.
The Pauli principle is crucial for the binding and structure of 6He, because it does not
only eliminate forbidden states between the core and the valence neutrons and between
the valence neutrons themselves, but it also delivers extra binding to the system through
exchange effects. Through explicit antisymmetrization done in MiCH we have found,
that the lowest hyper-spherical K = 0 three-body states are strongly blocked by the
Pauli principle between the core and the valence neutrons. This observation is important
especially in connection to three-body models, from which the hyper-spherical basis has
been adopted,and which employ different methods to account approximately for the core-
valence Pauli principle. The message here is that all three-body hyper-spherical states,
but especially those with K = 0, ought to be subjected carefully to the Pauli principle
if one is to describe 6He realistically. It is through the core—valence exchange effects that
the nucleus becomes three-body bound.
For a variationally Optimized 6He, the binding energy relative to the three-body
threshold, rms radii, the thickness of the neutron halo, and nucleon densities were com-
puted and found to agree with experimental values and results obtained in a variety of
structure models. The halo nature of the nucleus can be seen from its extended neutron
density resulting in the large difference between the matter and proton radius. It seems
that commonly computed and highly integrated observables such the three-body binding
energy and radii may not appreciate the amount of details built into different models of
6He. For these observables, a three-body approach with its simplistic description of the
4He core is as reliable as microscopic models.
To extract information about the 4He + n + n component in 6He, we have computed
the overlap integral between 6He and 4He. The integral was expanded in hyper-spherical
functions to make it comparable with three-body wavefunctions and applicable to re-
119
action calculations involving 6He. To our knowledge, this is the first time that overlap
functions for 6H8 have been calculated in this form from a microscopic structure model.
In agreement with three-body models, the microscopically founded three-body decom-
position from MiCH suggests that the same overlap channels dominate the 4He + n + n
cluster-division in 6He. On the other hand, MiCH predicts spectroscopic factors larger
by at least 30% than those from a three-body model for the dominant overlap channels
in 6He. This difference in spectrosc0pic factors reveals a deficiency of few-body models,
namely the inert-core approximation. Moreover, the relative enhancement of spectro-
scopic factors in MiCH varies between overlap channels, and so it may not be sufficient
in three-body models to simply renormalize the wavefunction to account for missing
microscopic information. In agreement with other models, we predict two major cluster-
ization patterns in 6He: the di-neutron-like and the cigar-like. In the former pattern, two
neutrons stay close together outside the 4He core, in the later pattern, two neutrons are
positioned on opposite sides of the core.
Finally, as a practical application of overlap functions obtained in MiCH, we car-
ried out a calculation for the reaction p(6He,4He)t at 25 MeV/ A assuming that the
reaction proceeds only through the simultaneous transfer of two neutrons. The angu-
lar distribution of the reaction cross-section is similar to that obtained with three-body
wavefunctions for 6He, but the cross—section with microscopically derived overlap func-
tions is larger by about 40% due to the above-mentioned difference in spectroscopic
factors between MiCH and three-body models. Even with microscopically derived input
for this reaction about the 4He + n + n component in 6He, theoretical predictions for the
cross—section assuming only the simultaneous two-neutron transfer do not reproduce ex-
perimental data. Therefore, we assert that other reaction mechanisms such as sequential
transfer and break-up should be included in theoretical considerations for this reaction.
120
6.2 Outlook
In the present work, the model MiCH has been formulated for bound states of two-
neutron halo nuclei and applied to the simplest case, 6He, bound by an effective soft—core
nucleon-nucleon interaction. For this nucleus, the model has proved to be working and the
original goal of combining advantages of few-body and microsc0pic models to describe
simultaneously the short-distance and the long-distance few-body halo correlations in
6He have been met. This success opens the door to possible future applications and
improvements of the model.
Even for 6He, there is still work to be done. An interesting application aimed on the
halo aspects would be the fi-decay of 6He to 6Li. There is experimental evidence that the
decay takes place essentially in the halo region in 6He. For such a study, a microscopic
wavefunction for 6Li is needed and we could attempt to produce it within MiCH despite
the fact that the ground state of 6Li does not have a Borromean character. 6Li would be
modelled as 4He + n + p with a microscopic 4He core.
The most exciting case to study among two-neutron halo nuclei is 11Li. This nucleus
has a very small two—neutron separation energy and a well developed neutron halo. As
part of the present work, 11Li has been studied within a deformed-core three-body model,
and it has been found that the core deformation plays an important role in the structure
of this nucleus. Given the physics insight built into MiCH, the model is well suited to
face the challenge of the unique neutron halo in 11Li. Going to mass eleven, however,
we would most likely encounter (serious) computational difficulties due to the memory
and CPU time required to carry out the Monte Carlo integration of matrix elements.
We believe that with the increasing computational power, improved algorithms and code
parallelization, this computational difficulty could be overcome which would bring 11Li
to our grasp.
A further improvement of the model could be achieved by the implementation of more
realistic nucleon-nucleon interactions and the inclusion of excited states. For light two—
neutron halo nuclei, including excited states means extension to the continuum, which is
121
important for reaction calculations involving these nuclei due to the proximity of break—up
thresholds. This brings us to the interplay between the structure and reactions.
In the present work, we have studied a two—neutron transfer reaction involving 6He
in the approximation of simultaneous transfer. Our results and those of other works
suggest that the experimental transfer data can not be reproduced unless other reaction
channels, such as sequential transfer and break-up, are included in the calculation. These
other channels involve continuum states of 6He as well as of 5H6. Our long-term goal is to
describe transfer reactions involving two-neutron halo nuclei in their full complexity with
a microscopically derived structure input for all nuclei involved in the reaction. This task
is important because it is at the intersection between nuclear structure and reactions,
where we have learnt most about two—neutron halo nuclei.
Finally, MiCH could be extended to study other light nuclei with less—straightforward
cluster divisions, but still showing few-body features. Among them are 8He with its
neutron-skin and 12C with its famous Hoyle state. For 8He (= 4He + n + n + n + n),
the hyper-spherical formalism in MiCH adopted from few-body models would be extended
to deal with the five-body inter-cluster motion, and for 12C (= 4He + 4He + 4He), the
two valence neutrons in the current version of MiCI-ll would be replaced by two microscopic
4He clusters. In both nuclei, all binary subsystems are unbound which makes MiCH well
suited to deal with such nuclei.
122
Appendix A
Implementation details
Along a random walk, local values of matrix elements in the spin-isospin space are com-
puted many times. This is done not only at each integration point, where local values
of operators such as 0106 in Eq. (4.22) and local weights such as w in Eq. (4.25) are
evaluated, but also at each trial move prOposed by the MetrOpolis algorithm, where a
local value of importance function needs to be calculated. In either case, the first thing
we need is a local representation of the wavefunction in Eq. (4.3).
A.1 Local representation of wavefunction
The concise coupled form of the wavefunction in Chapter 4 is not suitable for local ma—
nipulations. The reason is at least twofold. First, the angular momentum couplings hide
projections of orbital momenta along Jacobi coordinates, which are needed to evaluate the
spatial part of the wavefunction locally. Second, operators such as Hamiltonian contain
pair-wise (between two spots) operators acting on spin and isospin degrees of freedom in
the wavefunction, and so one must be able to identify spin-isospin ”values” of each spot.
A convenient basis of spin/—sospin states is provided by those sets in which each spot
has a definite third components of spin and isospin [140]. Therefore, at any set of spots
7" = {”1 , . . . ,FA}, angular momentum couplings in the wavefunction are decoupled so that
third components of orbital momenta along Jacobi coordinates and those of single-spot
123
spins and isospins can be identified. The difference between spots and particles has been
explained in Section 3.1.
In the system with A spots (particles), there are 2A possible permutations of spin
projections over spots. To represent a single spin in the computer, 0 is used fOr spin
”down” and 1 for spin ”up”. Each spin basis state in the wavefunction can then be
represented in a simple binary fashion. For example, for 4 particles we have:
spots —* 1 2 3 4
binary bits —> 3 2 1 0 binary #
0 0 0 0 0
0 0 0 1 1
0 0 1 0 2
111115
This basis only keeps a record of how spin projections are distributed over spots. Under
particle permutations imposed by Ava! and Acme—”“1 antisymmetrizers in Eq. (4.3),
particles carrying Spins jump between spots, but once the antisymmetrization of the
wavefunction has been completed, the only thing one needs to know is whether there is
a spin up or down at a given spot regardless of which particle brought it in.
The isospins can be handled similarly, but due to charge conservation, the number of
(2‘),
where Z is the number of protons. The number of isospin basis states could be reduced
isospin basis states can be reduced to:
further by constructing states with good total isospin, but then the action of isospin-
related Operators would be more involved.
After spatial parts of the wavefunction have been evaluated for each particle permu-
tation, all pieces of the antisymmetrized wavefunction belonging to a given spin-isospin
basis state are gathered. Ultimately, the wavefunction is represented locally as a two di-
124
mensional array of complex numbers with elements corresponding to different spin-isospin
basis states.
A.2 Operators
In the spin-isospin basis described in the previous section, the Hamiltonian containing
two-nucleon interactions is a sparse matrix, since two-body interactions can only change
the spins or isospins of two nucleons at a time. Any operator involving spins or isospins
is written in the form in which it acts directly on spin or isospin projections. Under the
action of such an operator, spin-isospin projections at some spots may be changed, and
the spin-isospin basis states are transformed among themselves. Spin-dependent operators
include for example interaction terms, the spin-orbit force, operators S2 and 82 of the
total spin and J2 and J; of the total angular momentum.
At the level of individual spots, a single spin Operator 8’ = éc‘f is expressed in terms
of raising 3+, lowering s‘ and oz operators:
1 3+ + s— 1 5+ — .9— 1
33; = 503; = —2—', 8y 2 §0y = T, S; = "2—0'z. (A.1)
When permitted, the z—projection of a single spin is raised and lowered upon the action
of 3+ and s‘, respectively. The spin projection is not changed by oz, but an extra
factor (—1) is acquired when the spin is down. An equivalent treatment is given to a
single isospin operator 5 = %'F. The interaction spin-exchange operator Pi‘; in Eq. (3.11)
simply swaps spin projections at spots 2' and j. The coordinate-exchange operator Pij
in Eq. (3.11) is written as Pi“;- = (—1)P,-‘;P,ZJ'-, where the isospin-exchange operator Pi;
swaps isospin projections at spots 1' and j. Sometimes, the central part of a two-nucleon
interaction is written in a form containing a,- - 0j and 7',- - 'rj operators instead of Pf; and
Pz-S. These Operators can be written as:
Ui-Uj=2P,g--1, Ti-Tj=2PZ:S-—1. (A.2)
125
Momentum-dependent Operators, such as the kinetic energy, the spin-orbit force, op-
erators L2 and L2 of the total orbital momentum and operators Jz, J2 of the total
angular momentum, involve first- and second-order derivatives of the wavefunction. All
required derivatives are obtained simply by moving each spot by a small distance in both
the positive and negative directions along each axis. If, for brevity, we consider a function
f (11:, y) of two variables a: and y, then its partial first- and second-order derivatives may
approximately be obtained from:
f(rc+A,y) — f(-”v— A41)
f5; = 2A , (A.3)
II f(+A1)+f( -A3)_2f(’)
rm: 2 a: y :2 y a: y ’ (AA)
II A7 A —A3 —A _ 1 II II
fry 2 f(a:+ 9+ )+f(::"l2 y ) 2f(rI=1/)_ m_ yy, (A5)
where A represents a small shift and the subscripts on f denote differentiating variables.
First- and second-order derivatives with respect to y would be obtained by making a small
shift in y instead of at. The dependence of derivatives on A must be tested carefully to
ensure their reliability. There should exist an interval of optimal values of A within which
the approximations to derivatives are A-independent. In MiCH, the value A = 0.001 fm
is used.
To obtain first- and second—order derivatives of the wavefunction in MiCH, Fq (A.3)~
Eq. (A5) are applied to each spot in each x, y, z direction which requires 2 X 3 x A ad-
ditional evaluations of the wavefunction. This is the reason why, from the computational
point of view, short but well decorrelated random walks may be much cheaper to work
with than long but highly correlated walks. Remember that when integration points are
being decorrelated, a single new evaluation of the wavefunction at Ftrz'al is needed in each
decorrleation step to check the Metropolis acceptance condition.
126
Appendix B
Further tests
For a success of any variational calculation in the framework of VMC, the reliability
and accuracy of matrix element evaluation are crucial. In Section 4.2.2, the Metropolis
algorithm has been checked for reliability, issues related to correlations inherent in the
algorithm were pointed out and methods of their suppression were outlined. In this Ap-
pendix, additional tests are presented to convince ourselves that numerical manipulations
and integrations in MiCH are carried out properly. Also, the problem of “bad points”
first mentioned in Section 4.2.3 is discussed
B.1 Triton tests and the story of bad points
This section summarizes some numerical results obtained within MiCH for the simplest
core + n + n bound nuclear system, the triton. In triton, the wavefunction in Eq. (4.3)
does not need to be core-valence antisymmetrized, because the core contains a single
proton distinguishable (by its isospin projection) from valence neutrons and there are no
core-valence forbidden states. This makes the triton a perfect case to check the imple-
mentation of the valence part of the wavefunction in Eq. (4.3).
We start with a triton containing a single valence term in the T Jacobi basis charac-
terized by {K , lx, ly, L, S, Jval, mag} and p0 from Section 2.1. By using the definition of a
local value of an operator from Eq. (4.22) and after applying the kinetic energy operator
127
from the left-hand side of Eq. (2.4) on triton’s wavefunction \I' from Eq. (4.3), the local
kinetic energy can be written as:
The“) =
(‘I’lTl‘m34 _ i 1
(‘qu’lsJ _ 2m P2115
_15 5_£’__1 P 2 5 P P 2 6 _ .0 2 7
[( 4 +214) 4(90) )Lnlag+(spo po Lnlag"1 po Lnlag‘z
52 K 32 K 52
+__( + /)g + /) (13.1)
2m p
with indexes on \11 omitted. Here, all L]‘(p/p0) are associated Laguerre polynomials having
the following explicit form taken from the relationship 22.3.9 in [97]:
L,((:::) =1§29) )m(n+k)—xm, 120/420 (B.2)
l
m=0 177.
By definition, we set Lil = LE2 E 0. For small hyper-radii, the local kinetic energy
diverges as:
K = 0: any nlag : Tloc(p _’ 0) z (1 + ”lag/3) /)0, (BB)
K 74 01 any "lag : Tloc(p _’ 0) z K(K + 4)/P2, (BA)
where the advantage was taken of the limiting form of Eq. (B.2). Furthermore, hyper—radii
in the triton are distributed according to:
1002) = p5 (7341.ng2, (8.5)
where ’leag is a hyper-radial basis function from Eq. (2.16) and p5 is the phase-factor
from the hyper-spherical volume element in Figure 2.1. By using Eq. (B.2), the hyper-
128
radial probability at small hyper-radii is:
n a 5 !
W —> 0) 4 1475-53-78 (8.6)
and it increases with the fifth power of may. At small hyper-radii, however, the local
kinetic energy in Eq. (8.3) and Eq. (B.4) diverges. Therefore, valence terms with higher
mag may trigger “bad points”, i.e. integration points with extraordinarily large local
kinetic energies. When that happens, the Monte Carlo energy averaging is harder to
converge.
To illustrate the effect of bad points, we consider a triton containing a single valence
term in the T Jacobi basis with l;c 2 lg = L = S = 0 and p0 = 0.50 fm. For the
hyper-spherical part of the wavefunction, two scenarios are considered: K = 0, may = 0
and K = 4, "lug = 2. The former case represents spatially the simplest possible triton,
whereas the later case contains non-trivial hyper-angular and hyper-radial parts. For
each of these tritons, a random walk is produced and the local kinetic energies and the
distribution of hyper-radii are shown in Figure B.1.
In the figure, our suspicion about bad points is clearly confirmed. Compared to the
hyper-radial distribution for the triton with K = 0, ”lug = 0 in Figure B.1(e), hyper-radii
for the triton with K = 4, ”log = 2 in Figure B.1(f) are shifted towards smaller values and
their probability grows rapidly near the origin, which gives birth to very large local kinetic
energies in Figure B.1(b) and Figure B.1(d) for that triton. Even though the number of
bad points is small, once they occur, they severely bias the Monte Carlo estimate of the
triton’s kinetic energy, sometimes making the estimate completely unreliable. A similar
effect has been observed in a fully antisymmetrized 6He, where inclusion of valence terms
with higher mag may trigger bad points, especially when there are only a few valence
terms in the wavefunction. In 6He, bad points may occur if any two spots and the
center-of-mass of remaining spots are close to each other. Then, in some core—valence
permutations, the valence part of the wavefunction is susceptible to large local kinetic
energies. Therefore, valence terms with may 91$ 0 should be added to the model space
129
o MiCH
theory
300 ' 1
g 200 - 1 %‘
Z a
8 8
'_- 100 7 1"—
0 7 a ‘
1 . 1 . 1 . 1 A 1 '2 A L+A+ 1 A A A
1° 0 2 4 6 8 10 12
9 [ml 911ml
(8) Triton with K = 0, may 2 0. I)ependence (b) Triton with K = 4, mag :2 2. Dependence
of local kinetic energies on the hyper-radius. of local kinetic energies on the hyper-radius.
10:594....” 105 .
a limo; EM,
; 4
'5 E ‘0 ’1
5 ‘ 3 5 103 1
'E ‘ ° 3
.9. 5 1 E. 102
E 5 3 2* l
= : 2 EE- 1
3 3 1° 1
'8 E '8 0 1
a 3 ‘3 1° II | I F
10'1 10“0
0 50 100 150 200 250 300 5000 10000 15000 20000
The [MeV] TIoc [MeV]
(c) Triton with K = 0, may = 0. Distribution of ((1) Triton with K = 4, may 2 2. Distribution
local kinetic energies. of local kinetic energies.
500. ' ' ' ‘ ' ‘ [:lMiCH 35°
'5- 400 g 300
'E » 3. 250
:- '9
19 30° ' s. 200
«1 .
g a
3.3 200 5 150
.o . g
g o 100
9 100 - ‘6.
a _ 50
0 00 4 8 12 16
9 [1m] 9 [fml
(e) 'Ii‘iton with K = 0, may 2 0. Distribution (f) "Ifiton with K = 4, may 2 2. Distribution
of hyper-radii. of hyper-radii.
Figure B.1: Local kinetic energies, their distribution, and the distribution of hyper—radii
for simple tritons with single K = 0, mag = 0 and K = 4, ”lug = 2 hyper-spherical
channels. Theoretical curves correspond to Eq. (B.1) for kinetic energy and Eq. (B.5)
for the hyper-radial probability.
130
of 6He only after a preliminary convergence in hyper-momentum has been reached for
may = 0.
It is gratifying to see that the theoretical predictions from Eq. (BI) and Eq. (8.5)
are perfectly reproduced by MiCH in Figure B.1. Finally, local kinetic energies can be
negative, as can be seen from Figure B.1(a) for the triton with K = 0, may = 0. When
the. wavefunction contains more valence terms, local kinetic energy can take negative
values even for small hyper-radii due to interference effects.
To put MiCH to a further test, we now try to reproduce basic observables of an
auxiliary triton produced within the three-body model described in Chapter 2. For this
purpose, we use the three-body code FaCE [93] to generate the wavefunction of a triton
within a limited model space containing all valence terms in the T Jacobi basis with
K g Kmax, may 3 10 and p0 = 0.50 fm. The nucleus is bound by the Volkov I interaction
with the mixing parameter set to m = 0.0 [135]:
Vi,- = —83.34 exp [—(r,-,-/1.60)2] + 144.83 exp [—(r,-,-/082)2] (13.7)
where rij 2 IF, — Fjl is the distance between nucleons z' and j. For several values of Kmax,
the binding energy and the rms radius of a triton produced by FaCE are estimated by
MiCH. The results are shown in Figure B.2. The binding energy of the triton predicted
by the three-body model is perfectly reproduced numerically by MiCH in Figure B.2(a).
For rms matter radii, there is a slight discrepancy (S, 1%) between the three—body and
MiCH values in Figure B.2(b). Even in the three-body calculations, some integrals are
carried out numerically with no error estimates provided. Therefore, the origin of the
mentioned discrepancy remains unknown. The actual difference between three-body and
our values may be beyond the accuracy of three-body calculations.
In summary, passing these triton and other tests, we have convinced ourselves that the
valence part of the wavefunction is implemented properly in MiCH. It is the valence part
of the wavefunction that is responsible for bad points plaguing Monte Carlo estimates of
matrix elements.
131
'7.4 I ' I ' I ' I 1 .75 *r ' I ' I V I . I I I
—-D—- FaCE
-7.6 _ o MiCH DID
. I— 1 1.74 r u—n/U
'7-8 - '1 H b /D/?/
5- .E.
0 -8.0 ' I! 1.73 ‘ ‘
2. FA
I“ -8.2 - « «$5
D\D ‘ 1.72 I ?_
-8.4 - \'\D~D——O—u—u 1 r —o- FaCE
[ . 1 . 1 . 1 . 1 . 1 ' 1 71 l A l A l A 1 . MiCH
'8'6 0 4 8 12 16 20 ' 0 4 8 12 16 20
Kmax Kmax
(8.) Binding energy. (b) Rms matter radius.
Figure B.2: The binding energy and the rms matter radius of a triton bound by the
Volkov I nucleon-nucleon interaction (m = 0). Empty squares are values calculated by
the three-body code FaCE. For Kmaa; = 0, 2, 4 and 10, the observables are estimated
by MiCH using wavefunctions produced by FaCE. Each value shown for MiCH is the
average value from eight independent random walks each containing 100,000 integration
points.
B.2 Additional tests and checks
Here, we briefly comment on some other tests of the accuracy of calculations in MiCH:
e For any wavefunction for the core produced within SVM as described in Chapter 3,
the binding energy is also estimated in MiCH by simply switching off the valence
part in Eq. (4.3). Our numerical estimates of the binding energy of the 4He core
are always in perfect agreement with values predicted by SVM.
o Occasionally, local values of the kinetic energy of the total center of mass are
computed at several integration points. Typically, the values for 6He are of the
order of 10"7 MeV, i.e. at the level of numerical noise.
0 At the beginning of each random walk, local values of the total orbital momentum
L = 2:421 I; and L2, total spin S = 2:le 3',- and S2, and total angular momentum
j = L + S and j2 are computed at a single integration point. Among them, only
local values of .12 and f2 are conserved. In 6He bound by central nucleon-nucleon
interactions, the core has L = S = J = 0 and the same is true for the valence part.
132
Then all operators have definite eigenvalues equal to 0. In such a case, typical local
values Of Operators are:
local value
operator
real part imaginary part
L1. -1.781E—16 207913107
Ly -2.664E—16 -2.282E—07
L... 2.559E-16 127313107
L2 1.300E-06 1.5215111
Sm 0.000E+00 0.000E+00
S, 0.00013+00 0.000E+00
S, 0.000E+00 0.000E+00
S2 5.0995131 -4.902E51
J.. -1.781E-16 2.079307
J, -2.664E—16 -2.282E-07
J2 2.559E—16 4.273507
.12 1.300E-06 1.5211911
Thus, all expected values are reproduced up to a numerical noise.
0 Also, at the beginning of each walk, the antisymmetry of the wavefunction is checked
at a single integration point. A properly antisymmetrized wavefunction III changes
its phase by (-1) when two particles are permuted under the action of the permu-
tation operator P. Therefore, the sum ‘1! + P‘IJ must vanish for any P. Again, this
is confirmed numerically up to numerical noise.
133
Appendix C
Comparative optimization on two
independent random walks
This Appendix presents the details of the improved optimization method employed
for 6He bound by central and non—central spin-orbit nucleon-nucleon interactions. The
method has been developed to circumvent numerical instabilities encountered for such a
6He due to the mixing of spin-singlet and spin-triplet valence terms in the wavefunction,
as mentioned in Section 4.2.3. The idea behind the method is to attenuate the effects of
statistical noise in the computation of overlap and energy matrix elements by comparing
results obtained on two independent random walks. In this section, energy means the
binding energy of 6He.
Let us suppose that somewhere along the optimization route, we have produced a
new best guess for the “stable” wavefunction containing “stable” valence terms. Two
independent “reference” (random) walks are produced for this wavefunction and the
reference energy of 6He is estimated on each of them. Then, several (many) “trial” valence
terms are temporarily added to the stable wavefunction. Such a temporary wavefunction
is called a “trial” wavefunction. Normally, trial terms include all valence terms absent
among the stable terms up to some maximum values of the hyper-momentum K and the
order of hyper-radial valence functions mag. Actually, two identical trial wavefunctions
(called trial wavefunctions) are created, each linked to one reference walk. For a given
134
trial wavefunction, all computations will be done on the reference walk attached to it. In
the following, energies are determined via the energy matrix diagonalization in correlated
sampling on reference walks. “A gain” is a difference between energy and the reference
energy. Trial terms are thinned out as follows:
1. Estimate the gain for each trial wavefunction. Except pathological situations, these
gains represent the maximum possible gains due to all trial terms.
2. On each reference walk, determine gains for all possible wavefunctions containing
“all stable + single trial” terms. A trial term is removed from a trial wavefunction
if its gain is positive, or is small in absolute value relative to the maximum possible
gain from step 1.
3. Compare trial terms outstanding in both trial wavefunctions and remove those
terms not present in both wavefunctions. This condition is highly selective.
4. On each reference walk, compute “cumulative” gains due to a singlet, doublet,
triplet, . . .of remaining trial terms with largest (in absolute value) individual gains
from step 2. Due to interference effects, a cumulative gain is not equal to a sum of
individual gains from step 2. A trial term is removed whenever its addition makes
the cumulative gain increase by more than a factor 2—3 compared to its individual
gain from step 2; large contribution to the cumulative gain could be an interference
effect or a numerical instability, two effects hard to disentangle.
5. Apply point 3 again on remaining trial terms.
6. At this point, there should be only few (5, 10) trial terms left and they are the same
in both trial wavefunctions. The number of remaining trial terms is controlled by
restrictions in steps 2 and 4. On each reference walk, find a trial term lowering
the energy the most, a pair of trial terms lowering the energy the most among all
pairs of trial terms, and so on for triplets, quadruplets, etc. When taken as absolute
values, the gain due to the most contributing singlet, doublet, . . .of trial terms is an
135
increasing function. The combination of trial terms for which the gain (in absolute
value) begins to saturate can be finally accepted to the stable wavefunction. The
selected final combination of trial terms must be the same on both reference walks,
though, which is normally true; when it is not true, all trial terms remaining after
step 5 can simply be accepted because they are guaranteed to be the same on both
reference walks. If the number of trial terms about to be admitted to the stable
wavefunction is too high, we restrict the pool of remaining trial terms by imposing
stricter conditions in steps 2 and 4.
Note that up to this point, we are only concerned about energy eigenvalues, which
are numerically stable.
7. In the previous step, a winning set of several trial terms has been found. These trial
terms came as winners on two independent reference walks. Therefore, their selec-
tion should be barely affected by possible numerical instabilities in the computation
of overlap and energy matrix elements and the energy matrix diagonalization. The
final energy matrix diagonalization for the system containing all stable + selected
trial terms provides numerically stable lowest eigenvalues and eigenvectors of linear
expansion coefficients in the wavefunction. The winning set of trial terms can be
safely accepted to the family of stable terms and a new stable wavefunction is thus
obtained.
However, the eigenenergies and eigenvectors in the last mentioned diagonalization
will most certainly differ on the two reference walks, which implies that two new sta-
ble wavefunctions are actually produced. The final check involves cross-correlated
runs, in which energy of a new stable wavefunction obtained on one reference walk
is computed in correlated sampling on the other reference walk and vice versa. If
everything is OK, linear coefficients in the two new stable wavefunctions can be av-
eraged coefficient by coefficient and the final new stable wavefunction is constructed
by using the averaged linear coeflicients.
As described in Section 4.2.3, several optimization cycles can be executed on the same
136
pair of reference walks. It needs to be emphasized that any failed trial valence term from
one cycle will appear on the list of trial terms in any subsequent cycle. Again, it is wise
to first build a wavefunction with several mag = 0 terms before higher-order Laguerre
polynomials are considered as trial terms. As optimization process progresses, the pool
of trial terms is broadened by an addition of terms with higher hyper-momentum and
mag. The Optimization continues until the convergence in the binding energy is reached.
In the present work, the non-linear parameter p0 is the same in all valence terms.
The optimization procedure described above assumes a constant value of po. Therefore,
the Optimal value of the non-linear parameter corresponding to the energy minimum still
needs to be found. An attempt to localize the global energy minimum by simply changing
the non-linear parameter in the converged wavefunction (accompanied by the energy
matrix rediagonalization) is doomed to fail. Such search would point to a fake energy
minimum formed at or close to the value of po, for which the wavefunction was originally
constructed. This is because due to the competitive selection, many valence terms have
not been admitted to the wavefunction, which makes the converged wavefunction firmly
tailored to a given non-linear parameter. This is to be compared with the case of central
forces, where all possible valence terms up to maximum values of K and ”tag are present
in any converged wavefunction, as explained in Section 4.2.3. To localize the global energy
minimum, we have to use a wavefunction containing all possible valence terms present
and absent in the originally optimized wavefunction. Once the global energy minimum is
found, the entire optimization process outlined above must be repeated for the optimum
value of pg.
137
Appendix D
Wavefunction normalization
In this Appendix, we discuss how to calculate the normalization of the 6He wavefunction.
In the present work, the normalization of 6He is needed to compute the overlap integral
between 4He and 6He in Section 5.3 in a meaningful way.
We rely on the Monte Carlo formalism developed in Section 4.2.1 and assume that
the wavefunction of a nucleus depends on all spatial, spin and isospin degrees of freedom,
i.e. \II = \I'(7"’, s, t). By using the Monte Carlo estimator from Eq. (4.14), the norm of the
wavefunction can be estimated as:
where w is a local weight:
w(T—.o(n)) _ (\Pl‘p>8,t
_ . D2
190" ("0 ( )
The sampling function p(f') satisfies Eq. (4.10) and can be chosen as a square of an
auxiliary sampling wavefunction ‘I’samplingffl depending on spatial coordinates only:
p(fl : wgampling(fl' (D3)
The sampling wavefunction should span the integration space of \IJ as closely as possible,
and yet the norm (‘1! samplingl‘I’ sampling) must be known analytically to be equal to one.
138
For compact nuclei such as 4He, the norm (\II|\II) can be estimated accurately by using
a simple Gaussian as the sampling wavefunction:
A
wgggggmg = Cexp £8 2 (F, — 18,-)? , (D.4)
i